Embed Size (px)
Citation preview
CER
N-T
HES
IS-2
015-
090
30/0
6/20
15
AALTO UNIVERSITY SCHOOL OF ENGINEERING Department of Applied Mechanics
Deepak Paudel
Quench Simulation of Superconducting
Magnets with Commercial Multi-Physics
Software
Thesis submitted in partial fulfilment of the requirements for the degree of
Master of Science in Technology
Espoo, June 30, 2015
Supervisor: Prof. Dr. Jani Romanoff
Instructor: Dr. Bernhard Auchmann
Abstract of master's thesis
2
Aalto University, P.O. BOX 11000, 00076 AALTO
www.aalto.fi
Author Deepak Paudel
Title of thesis Quench Simulation of Superconducting Magnets with Commercial Multi-Physics Software.
Degree programme Mechanical Engineering
Major Applied Mechanics Code K420-3
Thesis supervisor Prof. Dr. Jani Romanoff
Thesis advisor Dr. Bernhard Auchmann
Date 30.06.2015 Number of pages 62 + 39 Language English
Abstract
The simulation of quenches in superconducting magnets is a multiphysics problem of highest complexity.
Operated at 1.9 K above absolute zero, the material properties of superconductors and superfluid helium
vary by several orders of magnitude over a range of only 10 K. The heat transfer from metal to helium
goes through different transfer and boiling regimes as a function of temperature, heat flux, and transferred
energy. Electrical, magnetic, thermal, and fluid dynamic effects are intimately coupled, yet live on vastly
different time and spatial scales.
While the physical models may be the same in all cases, it is an open debate whether the user should
opt for commercial multiphysics software like ANSYS or COMSOL, write customized models based on
general purpose network solvers like SPICE, or implement the physics models and numerical solvers
entirely in custom software like the QP3, THEA, and ROXIE codes currently in use at the European
Organisation for Nuclear Research (CERN). Each approach has its strengths and limitations, some related
to performance, others to usability and maintainability, and others again to the flexibility of material
parameterizations. In this context the master thesis mainly involves the study of the strengths and
limitations of the first approach.
The primary goal of the thesis is to build a 1D numerical model representing a superconducting
wire based on existing physical models. An adiabatic model has been constructed, to solve one of the
five boundary value problems involved in the quench, both in ANSYS and in COMSOL. The temperature
dependent material properties and loads are defined using function tools in COMSOL and by creating
look up tables in ANSYS. The models were validated with QP3 and compared in terms of performance,
stability and accuracy.
The helium-cooled model is built only in ANSYS. The model solves two of the five boundary value
problems simultaneously as a coupled problem. Apart from generic numerical code (transient thermal
analysis), a separate algorithm is needed to define the non-linear heat transfer between the metal and the
helium. For this ANSYS Parametric Design Language (APDL) scripts are used. During the analysis the
ANSYS transient thermal codes are executed several times within a loop. There are three different types
of helium cooled models. All models were validated with QP3.
The results obtained from comparisons show that the adiabatic models were able to simulate
quenches with the desired accuracy. The adiabatic analysis in the commercial simulation tools is more
efficient and stable for various scale of spatial discretization. Similarly, the helium-cooled models are
able to simulate quenches with satisfactory accuracy. Nevertheless, the models are not compatible with
automatic time stepping method of the simulation environment. The use of fixed time stepping method
in the models resulted the coupled analysis in ANSYS to be far more time consuming than in QP3.
Keywords Superconductivity, Quench Simulation, Superconducting Magnets, Numerical Methods
3
Acknowledgments
Firstly, I would like to thank my university supervisor Jani Romanoff and CERN supervisor
Bernhard Auchmann to provide me the thesis opportunity, and for thoroughly checking the
thesis document despite of their busy schedule.
Thanks to Michal Maciejewski for helping me during the work as well as thoroughly
checking the thesis document. Thanks to my colleagues Ondrej Picha and Scott Rowan for your
sincere assistance during the work.
Finally, I would like to thank Arjan Verweij, Emmanuele Ravaioli, Valentina Venturi, Lauri
Kortelainen, Markus Bonda, Joel Grognuz, Ulf Friederichs and Pierre-Louis Ruffieux for
supporting me during the work.
4
Table of Contents
1. INTRODUCTION ................................................................................................................. 8
1.1 Superconductivity ............................................................................................................. 8
1.2 Superconducting Magnets ................................................................................................ 9
1.3 Quench Simulation ......................................................................................................... 11
1.4 Background to the research ............................................................................................ 11
2. QUENCH PROBLEM ........................................................................................................ 14
Copper Resistivity ............................................................................................................. 14
Critical current Ic(B,T) ...................................................................................................... 15
Current in the Copper-Matrix ........................................................................................... 15
Ohmic heating ................................................................................................................... 16
Helium thermal properties ................................................................................................ 17
2.1 Heat transfer from solid to helium .................................................................................. 17
2.1.1 Kapitza regime ......................................................................................................... 18
2.1.2 Film boiling-II regime .............................................................................................. 18
2.1.3 Natural Convection regime ...................................................................................... 19
2.1.4 Nucleate Boiling regime .......................................................................................... 19
2.1.5 Film-Boiling regime ................................................................................................. 19
2.2 Thermal problem in solids .............................................................................................. 20
3. NUMERICAL SOLUTION ................................................................................................ 22
3.1 The strong form .............................................................................................................. 24
3.2 The weak form ................................................................................................................ 25
3.3 Finite element approximation ......................................................................................... 26
3.4 Galerkin method to solve unknown degrees of freedom. ............................................... 28
3.5 Time Integration ............................................................................................................. 28
3.6 Non-Linearity ................................................................................................................. 29
4. SUPERCONDUCTING STRAND 1D MODEL ................................................................ 32
4.1 Material Properties ......................................................................................................... 33
4.2 Adiabatic Model ............................................................................................................. 34
4.2.1 Geometry and discretization .................................................................................... 34
4.2.2 Initial condition ........................................................................................................ 35
4.2.3 Loads and Boundaries .............................................................................................. 36
4.2.4 Solution and Results ................................................................................................ 36
5
4.3 Helium-cooled model ..................................................................................................... 41
4.3.1 Model 1: Switching of contact elements .................................................................. 42
4.3.2 Model 2: Updating heat transfer coefficient ............................................................ 44
4.3.3 Model 3: Updating the load ..................................................................................... 46
4.3.4 Solution and the results ............................................................................................ 47
4.4 Summary of the Results .................................................................................................. 53
4.4.1 Adiabatic model ....................................................................................................... 53
4.4.2 Helium-cooled model ............................................................................................... 53
4.4.3 Performance Analysis .............................................................................................. 53
5. CONCLUSIONS ................................................................................................................. 54
6. RECOMMENDATIONS FOR FUTURE WORK. ............................................................. 56
7. REFERENCES .................................................................................................................... 57
APPENDICES ......................................................................................................................... 63
1. Analytical Model .............................................................................................................. 63
2. Quench detection and Magnet protection ......................................................................... 63
3. Heat transfer in bulk helium ............................................................................................. 64
4. Non-linear inductance induced losses .............................................................................. 64
5. Resistivity of Cu ............................................................................................................... 65
6. Critical Current for Nb-Ti Superconductor ...................................................................... 65
7. Heat capacity of Nb-Ti ..................................................................................................... 66
8. Heat capacity of Cu .......................................................................................................... 67
9. Thermal conductivity of Cu .............................................................................................. 68
10. Thermal conductivity of He ............................................................................................ 69
11. Heat capacity of He ........................................................................................................ 70
12. Density of He .................................................................................................................. 71
13. 2D and 3D implication of 1D model .............................................................................. 72
14. APDL scripts: Adiabatic model ...................................................................................... 73
15. APDL scripts: Helium cooled model (Model 2) ............................................................ 77
16. APDL scripts: Helium cooled model (Model 1) ............................................................ 83
17. APDL scripts: Helium cooled model (Model 3) ............................................................ 87
18. APDL scripts: 2D quench simulation model of a magnet coil ....................................... 89
19. APDL scripts: 3D quench simulation model of a magnet coil ....................................... 95
6
Notation
aCu cross-section of copper fraction in a strand
aFB-I heat transfer coefficient, Film boiling-I regime
aFB-II heat transfer coefficient, Film boiling-II regime
aHe total contact surface area between helium and conductor
akap heat transfer coefficient, Kapitza regime
aNB heat transfer coefficient, Nucleate Boiling regime
aNC heat transfer coefficient, Natural Convection regime
awire cross-section of a strand
Bc critical magnetic field
Bc20 intercepts of critical surface along magnetic field axis
cp,cu specific heat capacity of copper
cp,model equivalent heat capacity of a model
cp,Nb-Ti specific heat capacity of Nb-Ti superconductor
cp,strand equivalent heat capacity of a strand
dCu equivalent diameter of copper fraction
dwire diameter of a strand
f copper to superconductor ratio in a Strand
fCu fraction of copper volume in a strand
hf heat transfer coefficient
hJoule joule heat generated per unit length in a strand
hHe rate of heat transfer to the liquid helium per unit area.
Ic critical current
INC current in the copper matrix
Jc critical current density
k thermal conductivity
L length of a strand
S heat generation load
Tb bath temperature
Tc critical temperature
Tc0 intercepts of critical surface along temperature axis
Tcs current sharing temperature
THe temperature of the liquid helium
Top operational temperature
Tpeak hot-spot temperature
Ts temperature of a strand
Tλ lambda temperature
ne empirical power constant
Qkap heat flux limit, Kapitza regime
QNC heat flux limit, Natural Convection regime
QNB heat flux limit, Nucleate Boiling regime
VHe total volume of helium around a strand
ρCu specific resistivity of copper
ρHe density of helium
7
8
1. INTRODUCTION
The main objective of the thesis is to build a quench simulation model of superconducting wire,
based on Finite Element Method (FEM), using general purpose commercial Multi-Physics
software packages, and to validate and compare the model with CERN in house software that
was specifically built for this purpose. In this chapter, we introduce some basic concepts of
quenches and quench simulation.
1.1 Superconductivity
Superconductivity is a unique property of certain materials that, when they are cooled down to
very low temperatures, exhibit zero resistivity. The very temperature below which the
conductor starts behaving like a superconductor is called critical temperature Tc. Besides critical
temperature there are two other parameters (i) critical current density Jc and (ii) critical
magnetic field Bc that determine whether the metal is in superconducting state or normal
conducting state. Assuming these parameters are the Cartesian coordinates, the region where
metal behaves as superconducting, and normal conducting outside the very region, can be
plotted. Figure 1 shows the critical surface of Nb-Ti alloy. Nb-Ti is commercially available and
widely used alloy in superconducting applications [1].
Figure 1 Critical surface for Nb-Ti [2].
9
1.2 Superconducting Magnets
To benefit from high efficiency and reduced size, superconducting magnets are used in various
applications such as magnets for nuclear fusion, Magnetic Resonance Imaging (MRI), MagLev
Magnetic Levitation (MagLev), and accelerator magnets [3 and 4].
The Large Hadron Collider (LHC) is a 27-kilometer-long chain of mostly
superconducting magnets which includes 1232 superconducting dipole magnets of length 15
metres to bend the beam of particles, 392 superconducting quadruple magnets of length
2.5-7 metres to focus the beam, and other more to steer the beam just before the collision[5].
The LHC ring is depicted in Figure 2. Two separate beams of particles are injected in opposite
direction and the collision of the particles takes place at particle detectors (A Toroidal LHC
Apparatus (ATLAS), Compact Muon Solenoid (CMS), A Large Ion Collider Experiment
(ALICE), and Large Hadron Collider beauty (LHCb)) where particle dynamics is studied. The
Radiofrequency (RF) cavities generate electromagnetic fields which produce forces to
accelerate the beam of particles once per turn. The cleaning sections in the ring consist of
structures called collimators. They scrape off off-center particles in order to protect the
accelerator against beam losses.
Figure 2 The Large Hadron Collider [5].
10
Superconducting wire also known as superconducting strand is the basic element of the
superconducting magnet coils. Figure 3e shows the Nb-Ti type superconducting strand where
tiny (~7 µm) Nb-Ti filaments are bundled together in a number of hexagons and embedded in
a copper matrix [1].
High-current Rutherford-type cable (Figure 3c) are used in LHC superconducting
magnets. Such cables are flattened helices usually comprising 20 to 40 strands. In order to
reduce cable eddy currents the strands are fully transposed with cable twist pitch of 6-8 times
the cable width which in turn improves the field homogeneity during the ramp [6].
The coil of LHC dipole magnets consists of two layers of Rutherford-type cable. The
cables are insulated to prevent shorts between the adjacent turns. The superconducting state is
obtained by immersing the magnet in a bath of liquid helium, which acts as a heat sink [7]. The
cross-sections of the coil along with the beam tube and the supporting structure is shown in the
Figure 3b.
Figure 3 a) Symbolic representation of the LHC ring. b) Cross-section of coil of superconducting dipole magnet.
c) Rutherford-type cable. d) Cross-section of Rutherford-type cable. e) Cross-section of a superconducting strand [1].
11
1.3 Quench Simulation
The transition of a conductor from the superconducting to the normalconducting state is called
quench. Typically a quench occurs because of sudden movement and friction-induced heating,
beam-induced losses and/or ramp-rate induced eddy-current losses; see Appendix 4.
In superconducting state, the resistivity of Nb-Ti filaments is zero and the entire current
flows through the filaments. During a quench, as the resistivity of the normalconducting
filaments is orders of magnitude higher than that of copper, the current deviates into the copper
matrix and generates Joule heating. The heat generated is taken away by liquid helium.
However if the rate of heat generation in the copper matrix persists to be greater than the rate
of heat transfer to the liquid helium the temperature of the conductor increases. Through heat
conduction it is transferred along the longitudinal direction of the strand and in the transverse
direction to the other strands thus successively raising the temperature beyond the critical
temperature and propagating the quench.
During the process the initial quench spot (so-called hot-spot) is exposed to high temperature.
The high temperature can lead to a meltdown of the insulation or even the copper matrix and
damage the magnet permanently. In order to protect the magnet, a quench needs to be detected
in time and protection schemes need to be implemented. For this, a proper understanding of the
quench behaviour is essential.
1.4 Background to the research
At CERN, research is being carried out to study the nature of elementary particles and forces
between them. The particles like protons are accelerated to nearly the speed of light, in the LHC
ring. When the beam of particles reaches the desired energy it is made to collide with another
beam travelling in the opposite direction. The experiments probe the basic nature of the most
fundamental particles.
Superconducting magnets in LHC are operated at cryogenic temperature; 1.9 K. Operating
magnets at such a low temperature, without a proper protection measures poses a threat of
magnet being permanently damaged by quenching. Simulation of quenches can help us
understand probable problems and how to resolve them before they occur in real world
situations.
CERN has its own in-house pieces of software for quench simulation such as Quench
Protection (QP3) [8], Thermal, Hydraulic and Electric Analysis (THEA) [9] and Routine for
the Optimization of magnet X-sections, Inverse field calculation and End design (ROXIE) [10].
12
QP3 is a FORTRAN code based on a 1D thermal network. It simulates the quench behaviour
of a longitudinally discretised conductor by an algorithm that goes through different loops and
iterations in order to solve the coupled and highly non-linear problem. THEA performs 1D/2D
static and transient multi-physics analysis of generic superconducting cable by thermal and
hydraulic networks. ROXIE is a 2D/3D simulation software for the design and optimization of
LHC superconducting magnets, featuring a magnet-level basic thermal network solver.
Similarly the use of multiphysics software like ANSYS and COMSOL [11, 12 and 13] and
network solvers like Simulation Program with Integrated Circuit Emphasis (SPICE) [14] has
been reported. ANSYS and COMSOL are examples of widely used Finite Element Analysis
(FEA) tools, whereas SPICE is a general purpose, open source network solver.
The Simulation of Transient Effects in Accelerator Magnets (STEAM) project, recently
started at CERN, aims to study the stability and performance of different approaches to quench
simulation in order to answer the following questions.
1. Can commercial multiphysics software deal with the complex physics models related
to superconductivity and superfluidity?
Commercial software has evolved over the decades and provides fast and accurate solutions for
a wide range of physical problems. However the codes may not include features to deal with
very specific problems such as superconductivity and superfluidity. Custom made software at
CERN was built specifically to solve such problems. Nevertheless in terms of performance they
might not be as efficient as compared to commercial software. The strengths and limitations of
both type of simulation environment should be compared by means of relevant benchmark.
2. Can all relevant physical effects be coupled?
In order to understand the quench behaviour of a magnet all involved physical phenomena have
to be treated as a coupled multiphysics problem. Most popular commercial software packages
have already incorporated the multiphysics features in their codes. For example, ANSYS has
two different methods to solve the coupled field problems.
I. Direct Method, where a coupled-field element is used that contains all necessary
degrees of freedom and the corresponding stiffness matrix and the load vector takes
into account all necessary terms.
II. Load Transfer Methods, which involve two or more analyses coupled by applying
results from one analysis as a load to another analysis.
13
Can these methods couple all the involved physical effects and if yes how efficient they can
be?
3. Can time stepping and spatial discretization be adaptive per physical model and
geometric subdomain?
Physical problems encountered during quenching exhibit different spatial and temporal scales
of integration constants, which are not uniformly distributed in the space-time domain. In order
to take advantage of the efficient numerical discretization, it is recommended to use adaptive
methods per physical and/or geometrical subdomains. Are there any straight forward solutions
or predefined functions in commercial software to use such features?
4. How does commercial multiphysics software compare to custom-made software in
terms of stability, accuracy and performance for our particular problem?
For now it is still unknown whether a commercial multiphysics software can simulate the
specific problems like quench taking into account all involved physical effects. If yes how
efficient would they be in terms of performance compared to custom-made software?
The thesis project is a part of the STEAM project which studies the coupling of different
commercial and in-house programs in a co-simulation framework. The result of the thesis
project shall give answers to questions 1 and 4 and contribute in providing necessary
information to prolong the research in finding answers to question 2 and 3.
14
2. QUENCH PROBLEM
Theoretical models that explain the physics behind the quench are already formulated and
documented in literature, e.g., [1, 6 and 7]. The simulation of quenches is a multiphysics
problem. Large magnets have multi-scale components from sizes as small as ~7 µm to 15 m in
length, with highly non-linear material properties that vary by several orders of magnitude over
a range of only 10 K. Moreover, various physical problems are entangled with each other
[15 and 16]. The coupled physical boundary value problems are:
A) The electrical problem: non-linear current-voltage characteristics of the superconductor;
non-linear dependency of the conductor resistivity on magnetic field, temperature.
B) The magnetic problem: non-linear inductance and eddy-current effects inside the coil
and in other structural elements.
C) The heat transfer from solid to helium: the heat transfer from the conductor to the helium
goes through different transfer and boiling regimes as a function of temperature, heat
flux, and transferred energy.
D) The thermal problem in solids: Joule losses in conductor, temperature dependent
thermal conductivity and heat capacity.
E) The thermal and fluid-dynamic problem of helium: temperature-dependent viscosity,
heat capacity, density, and thermal conductivity.
Among above five different boundary value problems the thesis project covers C and D.
D depends upon the current and magnetic field distribution obtained from A and B. Similarly
the heat transfer between the conductor and helium, C strongly depends upon E.
To make the problem simple the current and magnetic field will be considered to remain
constant throughout the analysis. As a first step only D will be solved, i.e, the adiabatic model
presented in Chapter 4.2. Then C and D will be solved as a coupled problem, i.e, the
helium-cooled model presented in Chapter 4.3.
Since C and D are strongly depended on A, B and E. It is essential to understand the
relation between these problems. In the following sections we present the related topics. In
Sections 2.1 and 2.2 we elaborate C and D respectively.
Copper Resistivity
The specific resistivity of the copper depends upon the temperature T and magnetic field B;
see Appendix 5. Residual Resistivity Ratio (RRR) is the ratio of the resistivity of the material
at room temperature and the cryogenic temperature (practically defined as 4 K) [4]. It is
15
determined by the purity of the copper. The higher the RRR value, the lower will be the
resistivity.
Critical current Ic(B,T)
The critical current in the superconductor depends upon the magnetic field B and the
temperature T. The critical current for Nb-Ti superconductor is fitted by [17]
𝐼c(𝐵, 𝑇) = (C1 + C2 ∙ 𝐵)(1 −𝑇
𝑇c(𝐵)). [A] (2.1)
The magnetic field direction is perpendicular to the current. The constants C1 = 3449 A and
C2 = -257 AT-1 for LHC type 01 cable are given in and Tc (B) is the critical temperature defined
as [4]
𝑇c (𝐵) = 𝑇c0(1 −𝐵
𝐵c20)0.59, [K] (2.2)
where Tc0 = 9.2 K and Bc20 = 14.5 T for Nb-Ti type superconductor; see Figure 5.
Current in the Copper-Matrix
Suppose the bath temperature in the magnet is Tb. Using (2.1) and (2.2) the critical current at
bath temperature is
𝐼c(𝐵, 𝑇b) = (C1 + C2 ∙ 𝐵)(1 −𝑇b
𝑇c(𝐵)), [A] (2.3)
using (2.1) and (2.3) critical current at temperature T is
𝐼c(𝐵, 𝑇) = 𝐼c(𝐵, 𝑇b)𝑇c−𝑇
𝑇c−𝑇b, 𝑇 < 𝑇c. [A] (2.4)
As one can notice from (2.4) when the temperature T in the conductor increases, the critical
current Ic (B, T) decreases. When it drops below the transport current I, the excess current starts
to commute to the copper matrix. The temperature at which the current starts to commute to
copper matrix is known as current sharing temperature Tcs. The current sharing temperature for
the magnetic field B and current I can be calculated as [10]
𝑇cs(𝐼, 𝐵) = 𝑇c(𝐵) (1 −𝐼
(C1+C2𝐵)). [K] (2.5)
A plot of critical surface showing the working line, Tc0, Bc20, and Tcs is shown in Figure 4.
16
Figure 4 Critical surface of Nb-Ti superconductor showing Tc0, Bc0, and Tcs for two different working points along the
load line [2].
The current in the copper matrix is equal to the transport current when the temperature of the
conductor rises beyond the critical temperature Tc. Following the above assumptions the current
in the copper matrix can be calculated using the empirical formula [10].
𝐼NC(𝑇) =
0 for 𝑇 < 𝑇cs𝐼 − 𝐼c for 𝑇cs < 𝑇 < 𝑇c𝐼 for 𝑇 > 𝑇c
. [A] (2.6)
Ohmic heating
Ohmic heating or Joule heating is the process in which heat is generated in a conductor as the
current passes through it. Assuming is the current density vector at a point in the conductor
where the temperature- and the field-dependent specific resistivity is ρ (T, B). The Joule heating
at the very point can be expressed as [10]
ℎJoule(𝑇, 𝐵) = 𝜌(𝑇, 𝐵) ∙ . [Wm-3] (2.7)
Assuming a Nb-Ti superconducting wire to be a one dimensional line with constant copper
cross-section area of acu, the Joule heat generated per unit length of the wire can be calculated
using (2.6) and (2.7)[10],
ℎJoule(𝑇, 𝐵) =
0 for 𝑇 < 𝑇cs
𝜌cu(𝑇, 𝐵)(𝐼−𝐼c)
2
𝑎cu2 for 𝑇cs < 𝑇 < 𝑇c
𝜌cu(𝑇, 𝐵)𝐼2
𝑎cu2 for 𝑇 > 𝑇c
, [Wm-3] (2.8)
17
where ρcu (T, B) is the resistivity of copper, Tcs and Tc are the current-sharing temperature and
critical temperature of the superconductor respectively.
Helium thermal properties
Helium is the only substance that remains liquid near the absolute zero temperature and does
not solidify below the pressure of 2.5 MPa. It has two liquid phases separated by a lambda line,
Tλ ~ 2.16 K; see Figure 5. Left to the lambda line it exists as superfluid liquid (He-II), i.e. it has
no measurable flow resistance (viscosity), and right to the line it exists as a normal fluid (He-I)
with measurable viscosity comparable to air [1, 3, 19, and 20].
The operating condition of the LHC superconducting magnets is indicated by a black dotted
line in Figure 5. The pressure remains near constant and He-II and He-I phases of helium are
used for cooling. As the temperature increases from 1.9 K to 4.5 K, helium goes through phase
changes (from He-II to He-I, and gaseous He) with highly non-linear behavior. The density of
the liquid decreases and there is a significant change in the thermal capacity and the
conductivity of the liquid; see Appendix 10 and 11.
Figure 5 Helium Phase Diagram [19]
2.1 Heat transfer from solid to helium
The heat transfer between the conductor and helium can be explained by assuming altogether
five different regimes and two different phases; see Figure 6.
18
Figure 6 Different assumed regimes to explain the non-linear heat transfer between conductor and helium.
He-II has unique cooling properties. The specific heat capacity and the thermal
conductivity in this phase are orders of magnitude higher than that of the electrical conductor.
There are two regimes explaining the heat transfer between He-II and the conductor.
2.1.1 Kapitza regime
Below a certain limit of heat flux, Qkap the cooling is dominated by Kapitza conductance. The
phenomenon provides the most effective heat flow. In general if two domains are in contact
there is a smooth temperature change in the interface as we go from one domain to another.
However between the conductor and He-II there exists sizable temperature jump. Details can
be found in [1]. The heat flow in this regime is given as
ℎ(𝑇s, 𝑇He) = akap(𝑇snkap − 𝑇He
nkap), [Wm-2] (2.9)
where Ts is the temperature of the conductor, THe is the temperature of the helium. akap and nkap
are the heat transfer coefficient (HTC) and the power coefficient, respectively. Typical values
are 200 Wm-2 Knkap and 4, respectively. The heat flux limit Qkap typically used is 35000 Wm-2.
2.1.2 Film boiling-II regime
Beyond the limit of heat flux Qkap a thin layer of helium vapor and/or liquid He-I is formed
between the surface of the conductor and the He-II. It prevents He-II to be in direct contact with
the conductor surface. The heat transfer is much less effective because of the low thermal
conductivity of the layer formed. If the heat flux drops below the limit, the cooling behavior
falls back to the Kapitza regime. The heat flow is expressed as
ℎ(𝑇s, 𝑇He) = aFB-II(𝑇s − 𝑇He). [Wm-2] (2.10)
The typical value of aFB-II is 250 Wm-2K-1.
He-I is a normal liquid with small thermal conductivity and large specific heat capacity.
This suggests that the heat transfer in this state is governed by convection and boiling rather
than by conduction. The He-I cooling can be modeled by three different regimes.
He-II
Kapitza
Film-Boiling-II
He-I
Natural convection
Nucleate boiling
Film-Boiling-I
19
2.1.3 Natural Convection regime
Below a few Wm-2 of heat flux there is no phase change and cooling is assumed to be purely
by convection,
ℎ(𝑇s, 𝑇He) = aNC(𝑇s − 𝑇He). [Wm-2] (2.11)
The typical value of aNC is 500 Wm-2K-1. The regime is assumed to continue until the heat flux
limit of QNC. The typical value of QNC is 10 Wm-2.
2.1.4 Nucleate Boiling regime
As the heat flux crosses the limit of pure convection, helium vapor is assumed to form at
preferred sites on the surface of the conductor. On increasing the heat flux the nucleation sites
get fully activated and form bubbles. The rate of bubble growth increases with increase in the
heat flux. As the bubbles detach from the surface, the cold liquid rushes down to cool the
surface. The following expression gives the amount of heat flow,
ℎ(𝑇s, 𝑇He) = aNB(𝑇s − 𝑇He)2.5. [Wm-2] (2.12)
The typical value of aNB is 50,000 Wm-2K-2.5. The heat flux limit QNB, typically used for this
regime is 10,000 Wm-2.
2.1.5 Film-Boiling regime
At still higher heat fluxes, the rate of bubble detachment increases. The bubbles become
unstable and they form a layer of helium vapor at the interface preventing He-I from being in
direct contact with the conductor. The heat flow is given by
ℎ(𝑇s, 𝑇He) = aFB-I(𝑇s − 𝑇He). [Wm-2] (2.13)
The typical value of aFB-I is 250 Wm-2K-1. Upon decrease in the heat flux, there is the possibility
to fall back to nucleate boiling and natural convection regime through a hysteretic effect. The
hysteresis is not taken into account in our models.
From the above assumptions an empirical formula to calculate the heat flow between the
conductor and the helium phases can be constructed,
ℎHe(𝑇s, 𝑇He) =
akap(𝑇s
nkap − 𝑇Henkap) for 𝑇He < 𝑇𝜆 & ℎHe < Qkap
aFB-II(𝑇s − 𝑇He) for 𝑇He < 𝑇𝜆 & ℎHe ≥ QkapaNC(𝑇s − 𝑇He) for 𝑇He ≥ 𝑇𝜆 & ℎHe < QNCaNB(𝑇s − 𝑇He)
2.5 for 𝑇He > 𝑇𝜆 & ℎHe > QNCaFB-I(𝑇s − 𝑇He) for 𝑇He > 𝑇𝜆 & ℎHe ≥ QNB
. [Wm-2] (2.14)
20
2.2 Thermal problem in solids
The main elements of the thermal problem in the superconductor are Joule heating and thermal
conduction. Joule heating depends upon temperature and magnetic field (2.7) and the thermal
properties are temperature-dependent; see Appendix 8 and 9.
Let T(X, t) the temperature at time t in point X of a conductor domain Ω with boundary, Γ;
see Figure 7. The change in the internal energy per unit time, at point X, inside the domain can
be given as [21]
Figure 7 Arbitrary domain subjected to internal heat source and surface flux distribution.
𝜕𝑈
𝜕𝑡= 𝜌cp
𝜕𝑇
𝜕𝑡, (2.15)
where ρ is the mass density and is assumed to be constant. cp is the temperature-dependent
specific heat capacity. The law of energy conservation states that the net energy flowing into
the domain per unit time must be equal to the change of internal energy inside the domain,
∫ 𝑆dΩ − ∫ 𝐧 ∙ 𝐪dΓΓ
= ∫ 𝜌𝑐p𝜕𝑇
𝜕𝑡ΩdΩ
Ω, (2.16)
where S is the temperature- and the field-dependent heat source distributed throughout the
domain. n is the normal vector and q the heat flux on the boundary. Using Gauss theorem on
the integral involving the surface flux,
∫ 𝑆dΩ − ∫ 𝛁 ∙ 𝐪dΩΩ
= ∫ 𝜌𝑐p𝜕𝑇
𝜕𝑡ΩdΩ
Ω, (2.17)
and hence
21
∫ [𝑆 − 𝛁 ∙ 𝐪 − 𝜌𝑐p𝜕𝑇
𝜕𝑡]dΩ = 0.
Ω (2.18)
Conservation of energy requires that (2.18) be true for all subdomains within the domain. This
yields
𝑆 − 𝛁 ∙ 𝐪 − 𝜌𝑐p𝜕𝑇
𝜕𝑡= 0. (2.19)
Fourier’s law gives the relation between heat flux and temperature gradient, i.e,
𝐪 = −𝑘𝛁T, (2.20)
where k is the temperature- and the field-dependent thermal conductivity. Using (2.19) and
(2.20) we get the second order heat balance equation [21],
𝜌𝑐p𝜕𝑇
𝜕𝑡= 𝛁 ∙ (𝑘𝛁T) + 𝑆. (2.21)
It is a governing differential equation to compute the thermal effects of the quench problem.
22
3. NUMERICAL SOLUTION
From a numerical-analysis point of view, quench simulations is one of the most interesting,
challenging and demanding subject in the superconducting magnet technology. Since the
1970’s various computational methods have been adopted to understand the quench behaviour
and successful 1D and 2D results have been reported for the strands and windings
[11-13 and 22-24]. However the simulation of large superconducting magnets is still a
challenging task even with today’s powerful computers and advanced computational methods.
We present four main aspects to be taken into consideration while solving the quench problem
numerically.
a) Transient problem
Quench simulation is a transient thermal problem. The simulation model seeks to solve the
temperature profile of the coil and its derivatives as a function of time. The accurate evaluation
requires the calculation of the temperature profile with smallest time steps in the order of a
micro-second.
b) Complex geometry of the coil
The geometry of the LHC superconducting magnet coil is designed based on a delicate
mathematical formulation [10]. The geometry construction for the numerical analysis using
such mathematical formulation is difficult. Considering the geometry to be a simple
(cylindrical/rectangular) shape does not simulate the problem with required accuracy. An
alternative approach is to import the geometry from a Computer Aided Design (CAD) system
into the simulation environment and simplify it for the analysis. An accurate model requires
construction of the actual up to 15 m long coil geometry with finite element discretization in
the scale of a millimeter.
c) Non-linear heat flux in the boundaries
The heat flux between the conductor and helium interface depends upon the temperature of the
conductor and the helium in contact, as well as on the heat-flux and integrated transferred
energy. A proper helium cooled model would take into account the phase change of the helium
and heat transfer through different boiling regimes; see Section 2.1.
d) Inhomogeneous and non-linear material properties
The coil is made of superconductor, copper and the insulation material, each having material
properties which exhibit a non-linear dependency on the temperature and the magnetic field.
The finite element method can easily take into account in-homogeneous material properties in
the calculation of the stiffness matrix. However, the non-linear problem usually involves some
23
form of iteration or drastically reduced time steps, and the process can be computationally
expensive. An efficient model would use adaptive temporal and spatial integration methods in
order to reduce the computation time.
The transient problem involving complex geometries, non-linear and inhomogeneous
material properties are discussed in the literature [25 and 26]. The popular commercial software
packages such as ANSYS, COMSOL, and ABAQUS have user friendly graphical user
interfaces to construct and solve such problems. However, modelling of the helium cooling in
the coil is still a challenging task. It is difficult to formulate the problem with coefficients
depending on two temperatures, heat flux and integrated heat. The traditional approach would
be to solve the temperature degree of freedom at each time step using numerical codes and to
introduce the cooling effect, based on the evaluated temperature degree of freedom of the
current time step, as a boundary load for the next time step. In other words coupling between
the generic numerical codes of the simulation environment with custom made cooling
algorithm. The methods are presented in Section 4.3.
In Chapter 2 we formulated the quench problem by deriving the governing heat diffusion
equation (2.21), the equation of joule heat generation (2.8) and the equation explaining heat
transfer between the helium and the superconducting wire (2.14). In Chapter 4 we will present
a numerical model of the superconducting wire, based on these equations, using commercial
software packages. These software packages are based on FEM. In this method the problem
domain is discretized into a number of small subdomains. The material constants, loads and
boundaries in each domain are expressed in matrix form. The equilibrium condition (balance
between internal forces corresponding to material properties and applied loads and boundaries)
is found by matrix manipulation. In the following sections (3.1 to 3.4), starting from the
governing heat diffusion equation (2.21), we present how such matrices are formed and
computed in the commercial simulation packages. The sections covers the basic theory of the
FEM, with a 1D superconducting wire as example. Furthermore, the sections 3.5 and 3.6
presents the numerical theory adopted by commercial software packages to solve the transient
and non-linear material problems respectively. In order to build an accurate, stable and efficient
numerical model it is essential to thoroughly understand the theoretical and practical
information provided in this chapter.
24
The steps involved in the FEM can be summarized as following [26 and 27]:
- Formulation of the problem in differential form along with boundary conditions. The
expression known as strong from.
- Transformation of the differential equation into weighted integral equations, integration
by parts to reduce the order of differential term in the integral equation and integration of
natural boundary condition in the integral form. The final expression known as the weak
form.
- Discretization of the domain into finite elements and approximation of the solution with
unknown degrees of freedom and shape functions.
- Expression of the equilibrium condition of all subdomains in matrix forms, using shape
functions and weighting functions, and solution of the unknown values.
3.1 The strong form
The strong form of the problem is the governing differential equation along with the necessary
boundary conditions. For a 1D superconducting strand the heat balance equation (2.21) reduces
to
𝜌𝑐p𝜕𝑇
𝜕𝑡=
∂
∂𝑥(𝑘
∂𝑇
∂𝑥) + 𝑆, 𝑥 ∈ Ω = [0, 𝐿], 𝑡 ∈ ψ = [0, 𝑡], (3.1)
where x is the spatial coordinate, L the length of the superconducting wire, and t the temporal
coordinate. The boundary condition is the restriction on the field variable or its derivatives on
the boundary. In our case temperature, T(x, t) is the field variable and its first derivative is
proportional to heat flux. The restriction the on field variable is called Dirichlet or essential
boundary condition and the restriction on the first derivative is called Neumann condition or
natural boundary condition. The spatial boundaries are the two ends of the wire. We assume
zero heat flux on both ends i.e. homogeneous Newman boundary conditions,
n ∙ q = n ∙ 𝑘𝜕𝑇
𝜕𝑥𝑒 = 0, 𝑥 ∈ 𝜕Ω = 0, 𝐿, (3.2)
where n is the outward pointing normal vector and 𝑒 is the direction of increasing x-coordinate.
The initial condition for the transient problem is
𝑇(𝑥, 𝑡) = 𝑔(𝑥) for 𝑡 = 0, (3.3)
where g(x) represents the temperature distribution on the wire at time t = 0.
25
3.2 The weak form
While solving differential equations analytically, a suitable function is introduced in the
differential equation and it is checked whether the function is a solution or not. To derive FEM,
a different approach is applied. Rather than searching for a solution fulfilling the above strong
formulation, it is required that the equation is fulfilled “in a weighted sense” for appropriately
chosen weighting functions ω(x). To facilitate the process, the strong form is changed to weak
form.
Multiplying (3.1) by a weighting function ω(x), rearranging and integrating over the spatial
domain, we get
∫ [𝜌𝑐p𝜕𝑇
𝜕𝑡𝜔 −
∂
∂𝑥(𝑘
∂𝑇
∂𝑥)𝜔
Ω− 𝑆𝜔]dΩ = 0. (3.4)
The higher the order of derivative term in the problem formulation the higher has to be the order
of polynomial functions of the approximation function. Integrating by parts reduces the order
of the derivative term in the integral form. We integrate by parts in the second term on the left
hand side of (3.4) to get
∫ 𝜌𝑐pΩ
𝜕𝑇
𝜕𝑡𝜔dΩ + ∫ 𝑘
∂𝑇
∂𝑥
∂𝜔
∂𝑥ΩdΩ − ∫
∂
∂𝑥(𝑘
∂𝑇
∂𝑥𝜔)dΩ
Ω− ∫ 𝑆𝜔dΩ
Ω= 0. (3.5)
By applying the fundamental theorem of calculus in the third term on the left hand side of (3.5)
we obtain
∫ 𝜌𝑐pΩ
𝜕𝑇
𝜕𝑡𝜔dΩ + ∫ 𝑘
∂𝑇
∂𝑥
∂𝜔
∂𝑥dΩ
Ω− ∫ n ∙ 𝑘
∂𝑇
∂𝑥𝑒𝜔dΓ
Γ− ∫ 𝑆𝜔dΩ
Ω= 0. (3.6)
Next, using the spatial boundary condition (3.2) in (3.6) and re-arranging the terms, the weak
formulation of the problem reads
∫ 𝜌𝑐pΩ
𝜕𝑇
𝜕𝑡𝜔dΩ + ∫ 𝑘
∂𝑇
∂𝑥
∂𝜔
∂𝑥dΩ
Ω= ∫ 𝑆𝜔dΩ
Ω. (3.7)
The highest order of derivative term present in the weak formulation (3.7) is the first order
derivative term. Therefore the solution of the problem has to be one time differentiable in the
weak sense i.e, it should be continuous. In addition it should satisfy the boundary conditions
(3.3). The simplest function or the lowest order function that could satisfy (3.7) and (3.3) is a
continuous piecewise function. Such function can have “kinks”, e.g, on material interface but
do not have any “jumps”.
26
3.3 Finite element approximation
The next step is to divide the spatial domain, Ω into n subdomains i.e. Ωi, i = 1, 2…, n (elements)
with total n+1 nodes at locations xi, i = 0, 1…, n (see Figure 8) such that the elements have
lengths xi – xi-1. The element-wise approximation functions are the polynomial functions of
order at least the order of the derivative term present in the weak form. We approximate the
temperature distribution in elements using a first order polynom Ti (x), i = 1, 2…, n.
𝑇𝑖(𝑥) = α𝑖,1 + α𝑖,2𝑥, 𝑥𝑖−1 ≤ 𝑥 ≤ 𝑥𝑖. (3.8)
Figure 8 Schematical view of domain discretization and linear solution approximation.
Using so called basis functions (3.8) we evaluate the temperature at the nodes of the element i.
We consider k = 1, 2 the local nodes of the element,
𝑇𝑖1 = α𝑖,1 + α𝑖,2𝑥𝑖−1, (3.9)
𝑇𝑖2 = α𝑖,1 + α𝑖,2𝑥𝑖. (3.10)
Rearranging (3.9) and (3.10) we get the constant terms,
α𝑖,1 = (𝑥𝑖𝑇𝑖1 − 𝑥𝑖−1𝑇𝑖2)/(𝑥𝑖 − 𝑥𝑖−1),
α𝑖,2 = (−𝑇𝑖1 + 𝑇𝑖2)/(𝑥𝑖 − 𝑥𝑖−1). (3.11)
Solving (3.11) with (3.8), we get
𝑇𝑖(𝑥) =𝑥𝑖−𝑥
𝑥𝑖−𝑥𝑖−1𝑇𝑖1 +
𝑥−𝑥𝑖−1
𝑥𝑖−𝑥𝑖−1𝑇𝑖2,
= ∑ 𝑁𝑖𝑘(𝑥)𝑇𝑖𝑘2𝑘=1 , (3.12)
27
where Nik (x), k = 1, 2, is the local element shape functions and is expressed as
𝑁𝑖𝑘(𝑥) =(𝑥−𝑥𝑖+1−𝑘)(−1)
𝑘
𝑥𝑖−𝑥𝑖−1. (3.13)
Figure 9 shows the shape functions, Nik (x) plotted using (3.13).
Figure 9 Linear shape functions Nik, the local nodes of the element i, k = 1, 2 (3.13).
The nodal shape function can be plotted in a global coordinate system based on the local
element shape functions. Figure 10 presents the nodal shape function φi(x) at location xi.
Figure 10 Nodal shape function φi (x) at location xi.
From Figure 10 we see
𝜑𝑖(𝑥) =
0𝑁𝑖2 𝑁(𝑖+1)10
=
0 for 0 ≤ 𝑥 ≤ 𝑥𝑖−1𝑥−𝑥𝑖−1
𝑥𝑖−𝑥𝑖−1 for 𝑥𝑖−1 ≤ 𝑥 ≤ 𝑥𝑖
𝑥𝑖+1−𝑥
𝑥𝑖+1−𝑥𝑖 for 𝑥𝑖 ≤ 𝑥 ≤ 𝑥𝑖+1
0 for 𝑥𝑖+1 ≤ 𝑥 ≤ 𝐿
, 𝑖 = 0, 1… , 𝑛. (3.14)
Suppose the temperature degrees of freedom at node xi and time t is Ti (t), the finite element
approximation of the temperature distribution throughout the spatial domain reads
28
𝑇ℎ(𝑥, 𝑡) = ∑ 𝜑𝑖(𝑥)𝑛𝑖=0 𝑇𝑖(𝑡). (3.15)
3.4 Galerkin method to solve unknown degrees of freedom.
The unknown degrees of freedom Ti (t) in the function (3.15) are solved by introducing the very
approximating function as the weighting function in the weak form. We choose the n + 1
(number of nodes) weighting functions φj, j = 0, 1…, n to be the nodal shape functions (3.15)
(the method known as Galerkin method) and rewrite the weak form (3.7),
∑ (∫ 𝜌𝑐pΩ𝜑𝑖𝜑𝑗dΩ)
d𝑇𝑖
d𝑡
𝑛𝑖=0 + ∑ (∫ 𝑘
d𝜑𝑖
d𝑥
d𝜑𝑗
d𝑥dΩ
Ω)𝑛
𝑖=0 = ∫ 𝑆𝜑𝑗dΩ,Ω ∀𝑗. (3.16)
The capacity matrix known as mass matrix and the thermal conductance matrix known as
stiffness matrix are computed for i, j = 0, 1, 2…, n respectively as
𝐂 = [C𝑖𝑗], C𝑖𝑗 = ∫ 𝜌𝑐pΩ𝜑𝑖𝜑𝑗dΩ,
𝐊 = [K𝑖𝑗], K𝑖𝑗 = ∫ 𝑘d𝜑𝑖
d𝑥
d𝜑𝑗
d𝑥dΩ
Ω. (3.17)
Similarly the heat generation load known as force vector is computed for i = 0, 1, 2 …, n as
𝐟 = [F𝑗], F𝑗 = ∫ 𝑆Ω
𝜑𝑗dΩ. (3.18)
The unknown scalar values known as displacement vector T = [Ti] for i = 0, 1, 2 …, n can now
be computed solving the simple equation system,
𝐂 + 𝐊𝐓 = 𝐟, (3.19)
with initial values T0 = [T (xi, 0)] = g(xi), i = 0, 1, 2…., n.
3.5 Time Integration
For transient problems the time domain ψ is considered to be infinite and we seek to find the
solution for a certain interval e.g. [0, t]. The interval is discretized into m sub-intervals with
time steps ti, i = 0, 1…, m corresponding to unknown scalar values of temperature degrees of
freedom, Ti, i = 0, 1…, m such that the time step size is Δti = ti – ti-1, i = 1, 2…, m. Assuming
the previous values of temperature degrees of freedom, Ti-1 are known, Ti is an unknown vector.
The approximation of the unknown temperature degrees of freedom and of its time derivative
is given respectively as
29
𝐓𝑖(𝑡) = 𝐓𝑖−1 +𝑡 − 𝑡𝑖−1
∆𝑡𝑖(𝐓𝑖 − 𝐓𝑖−1),
= 𝐓𝑖−1 + 𝜃(𝐓𝑖 − 𝐓𝑖−1), i = 1, 2…, m, (3.20)
and
𝑖(𝑡) =𝐓𝑖−𝐓𝑖−1
∆𝑡𝑖, i = 1, 2…, m, (3.21)
where θ = (t – ti-1) / Δti is known as weighting parameter. Updating (3.19) with (3.20) and (3.21)
we get
1
∆𝑡𝑖𝐂(𝐓𝑖 − 𝐓𝑖−1) + 𝐊[𝐓𝑖−1 + 𝜃(𝐓𝑖 − 𝐓𝑖−1)] + 𝐟 = 0, (3.22)
where 𝐟 = 𝐟𝑖 + 𝜃(𝐟𝑖 − 𝐟𝑖−1) is assumed to vary linearly over the time step. Rearranging (3.22)
gives the expression of the unknown vector [26],
𝐓𝑖 = (𝐂 + 𝜃∆𝑡𝑖𝐊)−1[(𝐂 − (1 − 𝜃)∆𝑡𝑖𝐊)𝐓𝑖−1 − ∆𝑡𝑖𝐟]. (3.23)
Different values of θ give different schemes of time stepping: θ = 0 (forward difference or
explicit Euler); θ = ½ (semi-implicit or Crank Nicholson); θ = 1 (backward difference or
implicit Euler) [28-30]. The commercial finite element packages usually use Newmark method
and General Hilber-Hughes-Taylor method for transient analysis, detail information can be
found in [31]. Adaptive time stepping methods are incorporated in some popular software
packages e.g. ANSYS, COMSOL. The quench simulation of large magnets needs adaptive
methods in order to reduce the computational cost.
3.6 Non-Linearity
In our case the specific heat capacity, thermal conductivity and the heat generation load have
non-linear dependency on temperature for which the system equation (3.19) takes the form,
𝐂(𝐓) + 𝐊(𝐓)𝐓 = 𝐟(𝐓). (3.24)
As can be seen from (3.24) the material properties change with the applied load. To solve such
problems an iterative method, the Newton-Raphson is used [32-34]. The method goes through
several iterations within a time step and updates the material matrices. The process is explained
in the following paragraph.
Provided the initial value of temperature degrees of freedom Ti-1 is given, the corresponding
capacity matrix Ci-1 and the stiffness matrix K i-1 will be always known. If the material
properties are assumed to vary linearly or remain constant over a time step, the equilibrium
30
equation (3.24) is not fulfilled throughout the time step. The amount by which the equilibrium
is out of balance is called residual vector given by
𝐑𝑗(𝐓) = 𝐟𝑗 − 𝐂𝑖−1,𝑗 + 𝐊𝑖−1,𝑗𝐓. (3.25)
Note: i represents time steps and j = 1, 2…, l represents iteration steps. The approximate
solution to (3.24) is obtained by repeating
i) Compute Ci-1, j, Ki-1, j and f j for Tj; for first iteration Tj = T1 =Ti-1,
ii) Improve the solution using Newton formula
𝐓𝑗+1 = 𝐓𝑗 − [𝐝𝐑𝑗(𝐓)
𝐝𝐓]−𝟏
𝐑𝑗(𝐓), (3.26)
until the absolute or relative L2(Euclidean) norm of the residual vector is below the specified
tolerance ϵ,
‖𝐑𝑗‖2 ≤ 𝜖, or ‖𝐓𝑗+1−𝐓𝑗‖2‖𝐓𝑗+1‖2
≤ 𝜖. (3.27)
31
32
4. SUPERCONDUCTING STRAND 1D MODEL
Superconducting strand is the basic element of the magnet coil. The simulation of a quench in
the strand helps to understand the overall picture of quench behavior and possible damage that
it can cause in the magnet. In order to detect a quench, the resistive voltage across the magnet
is measured and compared to a so called quench threshold; see Appendix 2. The results obtained
from the 1D model of the superconducting wire can be used to calculate the quench thresholds.
Increased accuracy of the model helps adjusting parameters of the quench protection system
which in turn contributes in the optimization of the magnet operation. Therefore so much
emphasis is given to build and improve the quench simulation model of the superconducting
wire.
In this chapter we explain the numerical model of the superconducting wire constructed
in commercial software packages. For this we choose two popular simulation environments i.e.
ANSYS and COMSOL. We take QP3 as reference to compare and validate our models. QP3 is
the FORTRAN code developed at CERN. The model represents superconducting wire as a
longitudinally discretized 1D network model. The code estimates the temperature profile of the
superconducting wire in the time domain. For more information see [8].
The numerical model simulates the quench behavior on the strand of the inner coil of the
LHC dipole magnets. The strand diameter is 1.065 ± 0.0025 mm. The strand has 8800 ± 20
number of Nb-Ti filaments of diameter 7 ± 1 μm. The copper to superconductor volume ratio
is 1.65 ± 0.03 [1].
The Nb-Ti filaments do not contribute to the electrical conduction after a quench, so heat
is produced only in the copper cross-section. On the other hand, since the temperature is
assumed to be homogeneous across the wire, the Nb-Ti fraction contributes to the overall heat
capacity. One could therefore build a model with actual diameter of the strand and change the
formula for heat production and heat conductivity or assume the model to contain only copper
fraction and change the formula for the heat capacity. We choose latter option. The diameter of
the assumed cylindrical strand domain, dCu is therefore smaller than the actual diameter of the
wire, dwire and is calculated as
𝑑Cu = √𝑓
1 + 𝑓 𝑑wire, [m] (4.1)
where f is the copper to superconductor volume ratio.
33
LHC dipole magnets are made up of Rutherford type cable. Figure 3d presents the
arrangement of the strands in the cable which suggests that the cylindrical surface of the strand
is not fully surrounded by liquid helium. We assume Vhe volume of helium surrounds the
conductor and the total contact surface area is aHe = s * L, where L is the length of the conductor
and s is the arc length parameter to assign an appropriate contact surface area between the
conductor and the helium. The cross-sectional properties corresponding to a particular strand
of the inner layer of the LHC dipole magnet’s coils and the values of the parameters used to
model the presence of helium around the strand is summarized in Table 1. The Vhe is chosen
for academic purpose, to enhance the effect of helium cooling. It does not reflect the actual
amount of helium present around the wire.
Table 1 Geometry parameters used for the quench simulation of 1D adiabatic model.
Parameter Symbol Value Unit
Length of the strand L 1 m
Diameter of the strand dwire 1.065e-3 m
Copper to superconductor ratio f 1.65 -
Total volume of helium VHe 1e-6 m3
Contact surface area s 1.320e-3 m
4.1 Material Properties
The thermal conductivity of the superconductor material is negligible as compared to that of
copper; hence is ignored [10]. The thermal conductivity of copper, kCu has a non-linear
dependency on the magnetic field and temperature; see Appendix 9. The heat capacity of the
copper, cp,Cu, and superconductor, cp,NbTi, is temperature-dependent; see Appendix 7 and 8. The
equivalent heat capacity of the strand is calculated as
𝑐p,strand(𝑇) = 𝑓Cu𝑐p,Cu(𝑇) + (1 − 𝑓Cu)𝑐p,NbTi(𝑇), [Jm-3] (4.2)
where fCu = f/(1+f) is the fraction of copper volume in the strand. Since the geometry of our
model compromises only the copper volume, the heat capacity to be defined to the model is
obtained as
𝑐p,model(𝑇) = 𝑐p,strand(𝑇)
𝑓Cu= 𝑐p,Cu(𝑇) +
1 − 𝑓Cu𝑓Cu
(𝑐p,NbTi(𝑇)),
= 𝑐p,Cu(𝑇) +1
𝑓𝑐p,NbTi(𝑇). [Jm-3] (4.3)
34
The thermal conductivity, density and the heat capacity of the helium are
temperature-dependent; see Appendix 10, 11 and 12.
COMSOL has so-called “function tools” that allow to define material properties as a
function of desired field variables. Similarly in ANSYS a table can be created with field variable
values in one column and the corresponding material property values in another column. Such
a function or a table is then assigned to the model such that the solver updates the material
property matrices every iterations based on the nodal values of field variables.
4.2 Adiabatic Model
In Chapter 2 we presented five different boundary value problems involved in the quench. In
this section we present a model constructed in COMSOL and in ANSYS to solve the one of the
problems, i.e the thermal problem in the solid. The model do not take into account the presence
of helium around the conductor and the heat diffusion is confined to the conductor domain only.
4.2.1 Geometry and discretization
The cylindrical conductor domain is discretised into n segments with n + 1 nodes as presented
in Figure 11.
Figure 11 Assumed geometry of the superconducting wire.
In COMSOL, a straight line was created via the Graphical User Interface (GUI) and the
“meshing tool” was used to discretise the domain. The cross-section properties were defined in
the domain definition section. We specify the problem type as a transient thermal analysis for
which the system equation to be calculated becomes equivalent to (3.24).
In ANSYS, dedicated APDL scripts were developed instead. Firstly, nodes were created
along a coordinate axis. Then, the nodes were connected with appropriate elements. We chose
the two-noded uniaxial thermal element FLUID116. It conducts heat between its nodes. The
element has two different types of degrees of freedom: temperature and pressure. We use only
the temperature degrees of freedom, for which the equilibrium equation of the elements
35
becomes equivalent to (3.24). ANSYS uses linear approximation function for this element; such
function is explained in Section 3.3.
The two main reasons why this element was selected are:
i) Material tables can be directly assigned to the element; see Section 4.1.
ii) It is a two noded element with possibility to apply surface loads.
4.2.2 Initial condition
We specify temperature at each node as an initial condition such that the temperature profile is
of Gaussian shape with a peak value at x = 0 m. Such a temperature distribution mimics the
initial heat pulse. It is calculated as
𝑇(𝑥) = Top + (Tpeak − Top)𝑒(−(
𝑥
𝛼)2), [K] (4.4)
where Top is the operating temperature and Tpeak is the hot-spot temperature, x is the node
position and α is the constant parameter that allows to adjust the shape of the Gaussian curve.
The curves in Figure 12 represent the initial temperature profile for two different values of α.
Figure 12 Gaussian type temperature profile representing the initial heat pulse.
The hot-spot temperature specified is well above the current-sharing temperature in order to
provoke a quench.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
2
4
6
8
10
12
14
16
18
20
Node position,x(m)
Tem
pera
ture
,T(K
)
= 0.1
= 0.05
36
4.2.3 Loads and Boundaries
For the adiabatic model the load is Joule heating given by expression (2.8). Our model is a
thermal model and considers only temperature as a field variable. The magnetic field and
current are introduced as constant values in order to reproduce conditions prior quench
detection. Hence, the equation of Joule heat generation is a function of one field variable, i.e,
temperature, T which is implemented as a table or a function.
The boundary condition in our case is zero heat flux on both ends. For 1D transient thermal
analysis it is the default setting in both COMSOL and ANSYS.
4.2.4 Solution and Results
The transient thermal analysis was carried out from 0 to t s. An automatic time stepping option
with minimum time step of 10 μs and maximum time step of 100 μs was specified to the solver.
Such scheme uses adaptive time stepping with time step size between specified minimum and
maximum value. To deal with material non-linearity the Newton-Raphson iteration scheme was
selected.
The current and magnetic field values selected corresponds to the strand of the inner coil
of the dipole magnet. Its location is indicated by black dot in the Figure 13. The strand being
close to beam is likely to quench due to beam induced losses.
Figure 13 Cross-section of the coil of LHC dipole magnets. Black dot indicate the location of the strand whose
cross-section properties, nominal current and magnetic field distribution are taken as a reference for the simulation.
Figure 14 represents the results obtained from COMSOL for the setting summarized in
Table 2. The value for Tpeak and α are arbitrary chosen and may not represent the actual
37
temperature distribution corresponding to the heat generated due to the induced losses or
frictional movement.
Table 2 Additional parameters used for the simulation of the 1D adiabatic model.
Parameter Symbol Value Unit
Discretization n 100 -
Transport current I 150 A
Magnetic field B 2.88 T
Operating temperature Top 1.9 K
Hotspot temperature Tpeak 20 K
Gaussian parameter α 1.5 m
Figure 14 Temperature distribution in the superconducting wire plotted for different time. The black dots depicts
advancing quench front.
Since the specified peak temperature is greater than the critical temperature, the quench always
initiates at the hot-spot. The quench may propagate further or vanish depending upon the
intensity of the heat pulse and the Joule heating corresponding to the transport current. On the
one hand the heat conduction phenomenon in the wire decreases the hot-spot temperature, on
the other hand the Joule effect increases the temperature of a node that is in the normal
conducting state. Since usually at the onset of a quench only a small fraction of strand is above
38
the critical temperature, in the beginning of the transient, the heat conduction phenomenon
becomes more effective compared to Joule heating and the hot-spot temperature drops. If the
intensity of heat pulse i.e. area covered by Gaussian curve is below a certain limit, the initial
drop in hot-spot temperature can fall below the critical temperature thereby preventing quench
propagation.
In our case one can notice from the figure that initially the hot-spot temperature drops
however does not fall below the critical temperature thus leading to quench zone expansion and
further heat generation. The generated heat propagates longitudinally successively raising the
temperature above critical temperature and propagating the quench.
The dotted horizontal line shown in Figure 14 represents calculated critical temperature
which is 8.073 K. We can calculate the quench propagation speed by plotting the position of
the quench front as a function of time i.e. the position of black dots in Figure 14 plotted against
the corresponding time; see Figure 15.
Figure 15 Quench front position as a function of time in reference to Figure 14.
The curves in Figure 16 present the results obtained from different simulation
environments at time t = 0.2 s for the settings summarized in Tables 1 and 2.
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.20
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
x-Q
uench f
ront
positio
n (
m)
Time (s)
Quench Speed = 4.50 m/s
39
Figure 16 Upper: Temperature profile of the superconducting wire estimated by adiabatic models of different
simulation environment, Lower: Difference in Temperatures profile in reference to QP3.
The red and black curves in the sub-plot is the calculated temperature difference between
ANSYS and QP3 and COMSOL and QP3 respectively. For the quenching strand it is likely to
have significant temperature difference at quench front position where there is rapid
temperature rise because of quench zone expansion. One can notice the phenomena from
Figure 16 at x = ~ 0.93 m. We are mainly interested in correctly estimating the hot-spot
temperature and quench propagation speed for which satisfactory results were obtained. Table 3
shows the difference in hot-spot temperature and quench propagation speed in reference to QP3.
Table 3 Variation in hot-spot temperature and quench speed in reference to QP3. Results corresponds to Figure 16.
Simulation Environment Hot-spot temperature difference
(K)
Quench propagation
speed difference (m/s)
ANSYS 0.18 (0.60%) 0.05 (1.09%)
COMSOL 0.11 (0.37%) 0.07 (1.53%)
To find out the appropriate scale of spatial discretization we did a convergence analysis
for the same settings summarised in Tables 1 and 2. The upper plot in Figure 17 presents the
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
5
10
15
20
25
30
Tem
pera
ture
,T(K
)
QP3
COMSOL
ANSYS
Initial T-Profile
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-2
0
2
Tem
pera
ture
err
or,
T(K
)
Node position, x (m)
QP3 - COMSOL
QP3 - ANSYS
Quench Speed (m/s)
QP3 = 4.57
COMSOL = 4.50
ANSYS = 4.62
40
quench propagation speed calculated for different mesh refinements, i.e, n values. The quench
propagation speed was calculated based on the initial and final positions of the quench front.
The minimum and maximum time stepping specified was 10 μs and 500 μs respectively.
Figure 17 Upper: Quench speed plotted for different mesh refinements, Lower: Change in quench speed with respect
to mesh refinement.
The lower plot calculates the difference in quench propagation speed between two adjacent n
values. The converged solution between two simulation environments varies by
0.114 m/s (2.49 %). At n = 1000 we reached the desired accuracy, i.e, ≤ 1 % change in quench
speed in both simulation environments. Therefore discretization in the scale less than or equal
to a millimetre is recommended for the adiabatic scenario of quench simulation.
Figure 18 compares the performance of different simulation environments in terms of
calculation time. The total computation time was plotted for different mesh refinements. The
Curves correspond to the convergence plot shown in Figure 17. The dashed lines indicates the
linear behaviour of the curves. The analysis were performed in an eight core 3.7 GHz PC
computer with 32 GB RAM (Random Access Memory).
102
103
104
105
4.5
4.55
4.6
4.65
4.7
4.75
Quench S
peed (
m/s
)
COMSOL
ANSYS
102
103
104
105
0
0.02
0.04
0.06
0.08
Q
uench S
peed(m
/s)
Discretisation, n
COMSOL
ANSYS
41
Figure 18 Total computation time plotted for different mesh refinements. The plotted results corresponds to the
convergence plot presented in Figure 17. Dashed lines indicates the linear behaviour of the curves.
One can notice from Figure 18 that for large number of elements ANSYS starts to perform
better. Since we need to discretise the strand in the scale less than or equal to a millimetre and
have to simulate much more than one meter of wire there will be more than 1000 elements in
the model for which the difference between COMSOL and ANSYS is of about 10%.
4.3 Helium-cooled model
In this section we present the models constructed in ANSYS in order to solve the two of the
five boundary value problems, i.e, A) the thermal problem in the conductor B) heat transfer
between the conductor and helium. The models solve the problems simultaneously as a coupled
physical problems.
The thermal problem in the conductor, A is already solved, i.e, the adiabatic model; see
Section 4.2. The heat transfer between the conductor domain (numerically the adiabatic model)
and the helium domain can be modeled with contact elements (representing convection,
radiation and/or conduction) on the interface; see Figure 19. ANSYS allows the heat transfer
coefficient (HTC) of such contact elements to be a function of one among following three
variables:
i) Temperature of one of the domain.
ii) Average temperature of both domains.
iii) Temperature difference between the domains.
102
103
104
105
101
102
103
Discretisation, n
Tota
l calc
ula
tion t
ime(s
)
COMSOL
ANSYS
QP3
42
With the above options it is not possible to implement the HTC that is equivalent to empirical
equation representing the heat transfer between conductor and helium (2.14). This shows that
the standard library of ANSYS is not generic enough to implement the complicated heat transfer
equation. Nevertheless, one can create a custom algorithm within ANSYS simulation platform,
using APDL scripts, and force ANSYS standard routine (transient thermal analysis) to run
several times within a loop such that every time steps appropriate heat transfer parameters, e.g,
HTCs are calculated and applied to the corresponding elements. In the following sections we
present three different approaches to update the heat transfer parameters. We distinguish them
as three different models.
4.3.1 Model 1: Switching of contact elements
The superconducting wire and the liquid helium are modelled as two parallel cylindrical
domains each discretised into n segments with n + 1 nodes as shown in Figure 19. The
dimension of the conductor domain is the same as in the adiabatic model. We use the thermal
element FLUID116 in both domains in order to calculate the heat conduction along the
longitudinal direction (3.24).
Figure 19 Model 1 helium cooled model.
By assigning material properties; heat capacity, density, thermal conductivity and the
viscosity to the helium domain it is possible to model the heat transfer phenomena within the
helium channel (bulk helium) which is in contrast with the heat transfer phenomena between
the conductor and the helium. The laminar (Landau) regime and the turbulent (Gorter-Mellink)
regime explains the heat transfer through the helium channel [19]; See Appendix 3. QP3
however does not consider heat transfer within the helium domain. To have a good agreement
in the result compared to QP3 the corresponding material properties of helium domain were
switched off.
The heat transfer between the domains is explained by 5 different regimes; see Section 2.1
Two different types of ANSYS contact elements were used, i.e, LINK31 to represent the
Kapitza regime and LINK34 to represent the remaining regimes.
43
I. Radiation Link - LINK31
LINK31 is a uniaxial two noded thermal element. It models the radiation heat flow rate between
its nodes. The equation is of the form
ℎ = σϵ(F𝑇𝑖4 − A𝑇𝑗
4), [Wm-2] (4.5)
where σ is the Stephan-Boltzmann constant, ϵ is the emissivity and F and A are additional
material constants. We assume σϵ = akap, F = Tinkap-4 and A = Tj
nkap-4 such that (4.5) becomes
equivalent to heat flow equation of the Kapitza regime (2.9).
II. Convection Link - LINK34
LINK34 is also a uniaxial two noded thermal element. It convects heat between its nodes. The
convection heat flow rate for this element reads
ℎ = ℎf(𝑇𝑖 − 𝑇𝑗)ne, [Wm-2] (4.6)
where hf is the HTC and ne is the empirical constant. Table 4 presents the assumed material
constants for the different regimes such that the equation (4.6) becomes equivalent to the
corresponding the heat flow equation.
Table 4 Contact elements and their corresponding real constants.
Regime HTC, hf(Wm-2K-ne) Empirical constant, ne
Film-Boiling II aFB-II 1
Natural Convection aNC 1
Nucleate Boiling aNB 2.5
Film-Boiling I aFB-I 1
The heat transfer surface area defined to the contact elements is equivalent to the total
contact area between the conductor and helium. In this model we have three different types of
elements, seven different types of materials and in total 2n + 5(n + 1) number of elements.
The imposed loads and the initial condition for the conductor domain, i.e, nodes
i = 1, 2…., n are the same as defined in the adiabatic model. The initial temperature defined to
the helium domain i.e. to nodes j = 1, 2…., n is equal to Top.
Figure 20 shows the block diagram of the algorithm built to impose a governing regime of
heat transfer during the analysis. The code can be found in the Appendix 16. The time steps are
represented as k = 1, 2…, m.
44
Figure 20 Block diagram of the algorithm that switches appropriate element every time steps.
To take into account all regimes of heat transfer we join 5 contact elements of different material
constants between every pair of nodes such that each contact element represents a single regime.
In the beginning of the analysis all contact elements are inactive. At every time step, the
temperatures at the nodes and the heat flow rates in the contact elements is extracted from the
solutions of the previous time step. Based on extracted temperature values and the heat flow
rates an applicable regime of the heat transfer is determined for each node pair and the
corresponding contact element is activated. The algorithm allows only one contact element to
be active at the same time. For each node pair the analysis starts with the lowest regime, stays
until the heat flux is above the limit, and shifts to another higher regime with a condition that it
can fall back if the heat flux drops below the limit.
4.3.2 Model 2: Updating heat transfer coefficient
The cylindrical conductor domain is discretized as in the adiabatic model. Its nodes are linked
to helium masses of volume Vhe/(n+1) as shown in Figure 21.
Figure 21 Model 2 helium cooled model.
45
The link is a contact element LINK34. One-noded mass elements MASS71 are used to model
the helium masses. In this scheme the heat conduction along the helium is not taken into
consideration. Unlike in the previous model only one contact element is connected between
each node pairs. There are altogether n + 2(n+1) elements in the model.
The load and the initial condition in the conductor domain and the initial condition in the
helium masses are the same as in the previous model (Model 1). The heat flow rate in the contact
element is of the form (4.6). We set the empirical constant ne = 1 and redefine the flow rate as
ℎ𝑘+1 = ℎf𝑘+1(𝑇𝑖
𝑘 , 𝑇𝑗𝑘)(𝑇𝑖
𝑘+1 − 𝑇𝑗𝑘+1), [Wm-2] (4.7)
where k = 1, 2…, m represents time steps. The (5.7) becomes equivalent to (2.14) with
ℎf𝑘+1(𝑇𝑖
𝑘 , 𝑇𝑗𝑘) =
akap
(𝑇𝑖
nkap, 𝑘−𝑇
𝑗
nkap, k)
𝑇𝑖𝑘−𝑇𝑗
𝑘 𝑇𝑗𝑘 < 𝑇𝜆 & ℎ
𝑘 < Qkap
aFB-II 𝑇𝑗𝑘 < 𝑇𝜆 & ℎ
𝑘 ≥ Qkap
aNC 𝑇𝑗𝑘 ≥ 𝑇𝜆 & ℎ
𝑘 < QNC
aNB(𝑇𝑖𝑘 − 𝑇𝑗
𝑘)1.5 𝑇𝑗
𝑘 > 𝑇𝜆 & ℎ𝑘 > QNC
aFB-I(𝑇𝑖𝑘 − 𝑇𝑗
𝑘) 𝑇𝑗𝑘 > 𝑇𝜆 & ℎ
𝑘 ≥ QNB
. [Wm-2] (4.8)
The algorithm to update the HTC of the contact elements is depicted in Figure 22.
Figure 22 Block diagram of the algorithm that calculates heat transfer coefficient of the contact elements every time
steps.
46
At every time step we extract the temperature at nodes and the heat flow rates in the contact
elements. Based on the values obtained we calculate HTC for each contact elements using (4.8).
4.3.3 Model 3: Updating the load
The model is similar to the adiabatic model. We do not represent the helium as a separate
domain, instead we introduce the helium cooling effect as negative heat generation load in the
equilibrium equation. The amount of heat flow to the liquid helium in contact is subtracted from
the Joule heat generated at nodes i = 1, 2…, n + 1. In Figure 23 the time steps are represented
as k = 1, 2…, m and j = 1, 2…, n+1 is used to define the temperature of liquid helium in contact
at locations corresponding to the nodes of the conductor domain.
Figure 23 Model 3 Helium cooled model.
Given the initial temperature, material properties and the volume of the liquid helium, the
temperature of the liquid helium in contact at several locations for different time steps is
expressed as
𝑇𝑗𝑘+1 =
ℎHe𝑘 𝑎He
𝜌He(𝑇𝑗𝑘)𝐶p,He(𝑇𝑗
𝑘)𝑉He∆𝑡 + 𝑇𝑗
𝑘, [K] (4.9)
where ρHe is the density, cp, he the heat capacity and Vhe the total volume of the liquid helium.
hkHe is the amount of heat flow to liquid helium, aHe the surface area of the conductor in contact
with liquid helium and Δt the time step size.
The initial temperature of the nodes in the conductor domain is the same as in the adiabatic
model and initial temperature of liquid helium in contact, Tj1 = Top. The net heat generation load
to be updated at every time step reads
𝑆𝑖𝑘+1(𝑇𝑖
𝑘, 𝑇𝑗𝑘+1) = ℎJoule
𝑘+1 (𝑇𝑖𝑘) −
𝑎He
𝑎Cu𝐿ℎHe𝑘+1(𝑇𝑖
𝑘, 𝑇𝑗𝑘+1), [Wm-3] (4.10)
where aCu is the cylindrical surface area calculated considering only copper volume and L is the
length of the conductor. The Joule heating function, hJoulek+1(Ti
k) for constant magnetic field can
47
be calculated using (2.8) and the rate of heat flow, hHek+1(Ti
k, Tik+1) can be calculated using
(2.14). Figure 24 presents the block diagram of the overall process of updating the load vector.
Figure 24 Block diagram of the algorithm that updates the load values to the elements every time steps.
For the initial time step there are no loads. For the rest of the steps, first the temperature of
liquid helium is updated then the load vector is updated based on the new temperature of liquid
helium, previous temperature of the superconductor and the previous heat flow rate to helium.
4.3.4 Solution and the results
For all of the above models (Model 1-3) transient thermal analysis was carried out from 0 to t s.
Since we need to update the heat transfer parameters every time step it is not possible to use the
adaptive time stepping scheme of ANSYS. Building custom time stepping algorithm was
beyond the scope of the thesis therefore we did not go further in programming ourselves and
used fixed time stepping scheme. The time step size equal to 10 μs was specified to the solver.
Similar to the adiabatic model, the Newton-Raphson iteration scheme was selected to deal with
the material non-linearity. Figure 25 shows the results obtained for the setting given in the
Table 5.
48
Table 5 Simulation parameters defined for the quench simulation in helium-cooled models.
Parameter Symbol Value Unit
Discretization n 1000 -
Transport current I 150 A
Magnetic field B 2.88 T
Operating temperature Top 1.9 K
Hotspot temperature Tpick 20 K
Gaussian parameter α 0.1 m
Total simulation time t 0.2 s
Empirical constant nkap 3 -
HTC (Kapitza regime) akap 200 Wm-2 Knkap
HTC (Film boiling II) aFB-II 200 Wm-2 K-1
HTC (Natural convection) aNC 500 Wm-2 K-1
HTC (Nucleate boiling) aNB 50000 Wm-2 K-2.5
HTC (Film boiling I) aFB-I 220 Wm-2 K-1
Heat flux limit (Kapitza regime) Qkap 35000 Wm-2
Heat flux limit (Natural convection) QNC 10 Wm-2
Heat flux limit (Nucleate boiling) QNC 20000 Wm-2
Figure 25 Contour plot of Temperatures (K) obtained for from the analysis for the setting summarised in Table 5,
Uppper: From Model 2, Lower: From Model 3.
Table 5 shows that the superconductor domain is discretised in the scale of millimetre. The
discretization is so fine that it is difficult to distinguish uniaxial contact elements and the result
is obtained as a 2D plot; see upper part of Figure 25. The lower edge of the 2D plot represents
the conductor domain and the upper edge represents the helium domain. The temperature
49
gradients seen in the contact elements is the linearized representation of the temperature
difference between two domains.
Figure 26 presents the temperature estimation on the superconducting wire at different
times. The curves were obtained from ANSYS Model 1. Similar results were obtained from
Model 2 and Model 3 as well. The red curve depicts the imposed initial temperature profile.
Figure 26 Temperature profiles of the superconducting wire, at different time, estimated by ANSYS Model 1, for the
setting summarised in the Table 5.
Similar to the adiabatic model the hot-spot temperature drops initially then as quench
propagates the temperature keep on rising until the end of analysis, nevertheless the quench
propagation is significantly slow in this case.
Figure 27 compares the temperature on the superconducting wire estimated by ANSYS
Models and QP3 for the setting summarised in Table 5. The blue curves represents the imposed
initial temperature distribution, and the red curves in the sub-plots is the calculated temperature
difference between ANSYS and QP3.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
5
10
15
20
25
Node position, x(m)
Tem
pera
ture
,T(K
)
0.000084 s
0.000172 s
0.000259 s
0.045131 s
0.064001 s
0.2 s
Initial T-Pulse
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.162
4
6
8
10
12
14
16
18
20
Length, L(m)
Tem
pera
ture
,T(K
)
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
5
10
15
20
25
Length, L(m)
Tem
pera
ture
,T(K
)
0.000084 s
0.000172 s
0.000259 s
0.045131 s
0.064001 s
0.2 s
Initial T-Pulse
50
Figure 27 Temperature profile of the superconducting wire estimated by ANSYS Models and QP3 at time t = 0.2 s.
The plots show satisfactory results in temperature estimation except the expected significant
difference at quench front position. The variation in the estimation of the hot-spot temperature
and the quench propagation speed is presented in Table 6.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
5
10
15
20
25
Te
mp
era
ture
,T(K
)
Model 1; Element Switching
QP3
ANSYS
Initial T-Profile
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-1
0
1
2
Te
mp
era
ture
err
or,
T(K
)
Node position, x (m)
QP3 - ANSYS
Quench Speed = 0.520 m/s
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
5
10
15
20
25
Te
mp
era
ture
,T(K
)
Model 2; Updating of HTC
QP3
ANSYS
Initial T-Profile
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-0.5
0
0.5
1
1.5
Te
mp
era
ture
err
or,
T(K
)
Node position, x (m)
QP3 - ANSYS
Quench Speed = 0.530 m/s
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
5
10
15
20
25
Te
mp
era
ture
,T(K
)
Model 3: Updating of Load
QP3
ANSYS
Initial T-Profile
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-1
0
1
2
Te
mp
era
ture
err
or,
T(K
)
Node position, x (m)
QP3 - ANSYS
Quench Speed = 0.525 m/s
51
Figure 28 shows the quench front position at different time. Curves are in reference to
plots in Figure 26 and similar plots obtained from other two ANSYS Models and QP3. The
quench speed is calculated based on the initial and final position of the quench front.
Figure 28 Quench front position with respect to time estimated by ANSYS models and QP3.
Table 6 Variation in the hot-spot temperature and quench propagation speed estimated by ANSYS Models in
reference to QP3.
Simulation Environment Hot-spot temperature
difference (K)
Quench propagation speed
difference (m/s)
Model 1 0.11 (0.46%) 0.015 (2.80%)
Model 2 0.11 (0.46%) 0.005 (0.93%)
Model 3 0.09 (0.37%) 0.010 (1.87%)
The hot-spot temperature and the quench propagation speed estimated by the models are
satisfactory. Table 6 shows that the Model 2 is more accurate among the three different models.
One can notice from Figure 16, 27 and 28 that the quench propagation speed estimated by the
helium cooled models is nearly ten times slower compared to adiabatic models, whereas, the
estimated rise in the hot-spot temperature is nearly half. This suggests that the helium cooling
is contributing more in slowing down quench propagation than in taking heat away from the
conductor.
The time step size varies depending upon the type of problem. For non-linear problems
time step size needs to be very small. We did convergence analysis to find out the minimum
time step size for our problem. The results are shown in Figure 29. For the same settings
presented in Table 5 the temperature profile at time 0.01 s is plotted for different time step sizes.
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.20.08
0.1
0.12
0.14
0.16
0.18
0.2
0.22
Time (s)
x-Q
uench f
ront
positio
n (
m)
QP3
Model 1
Model 2
Model 3
Quench Speed (m/s)
QP3 = 0.535
Model1 = 0.520
Model2 = 0.530
Model3 = 0.525
52
Figure 29 Convergence plot: Temperature profile of the superconducting wire at t = 0.01 s for different time step
sizes.
Figure 30 shows the hot-spot temperature (black curve) for different time step sizes. The
green curve is the calculated difference in hot-spot temperature between two adjacent time step
sizes.
Figure 30 Upper: hot-spot temperature for different time step sizes. Lower: Difference in hot-spot temperature
between two adjacent time step sizes.
For the time step size smaller than 10 μs, we achieved the desired accuracy i.e. ≤ 1% change in
the hot-spot temperature. The total time taken in ANSYS Model 2 to simulate the phenomena
for 0.001 s with time step size equal to 10 μs is 280 s, whereas in QP3 adaptive time stepping
method is used where the lower and upper bounds of the time step size are 3 μs and 100 μs
respectively and total time taken is 5.5 s.
10-6
10-5
10-4
14.5
15
15.5
16
16.5
Ho
t-sp
ot te
mp
era
ture
, T
(K)
10-6
10-5
10-4
0
0.5
1
1.5
T
(K)
Time step size, t(s)
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
2
4
6
8
10
12
14
16
18
20T
em
pe
ratu
re, T
(K)
Node position, x(m)
Initial T-Profile
t = 5e-5 s
t = 2e-5 s
t = 1e-5 s
t = 5e-6 s
t = 2e-6 s
0 0.02 0.04 0.06 0.08 0.12
4
6
8
10
12
14
16
53
4.4 Summary of the Results
4.4.1 Adiabatic model
Validation of ANSYS and COMSOL model with QP3 (Figure 16)
In reference to QP3, the error in the hot-spot temperature and quench propagation speed
estimated by ANSYS is 0.60% and 1.09% respectively. Similarly, the error by COMSOL is
0.37% and 1.53 % respectively. This shows that both COMSOL and ANSYS are able to
calculate hot-spot temperature and quench propagation speed with desired accuracy, i.e, < 1%
error in hot-spot temperature and < 2% error in quench propagation speed.
Convergence analysis with respect to mesh refinement (Figure 17).
For both ANSYS and COMSOL a converged solution was obtained for spatial discretization in
the scale less than or equal to a millimetre for which the change in quench speed dropped below
1%. The converged solution of ANSYS and COMSOL varies by 0.4 % in hot-spot temperature
and by 2.49% in quench propagation speed. The required scale of mesh refinement, i.e, a
millimetre scale is a well-known empirical value. The results obtained from the convergence
analysis is as expected.
4.4.2 Helium-cooled model
Validation of ANSYS model with QP3 (Figure 27, Figure 28)
In comparison to QP3, the error in hot-spot temperature and the quench speed estimated by the
most accurate ANSYS model is 0.46 % and 0.93% respectively. This shows that ANSYS is
able to calculate hot-spot temperature and quench propagation speed with relative accuracy.
Convergence analysis with respect to time step size (Figure 30)
A converged solution was obtained for time step size less than or equal to 10 μs for which the
change in hot-spot temperature dropped below 1%. The minimum required time step size, i.e,
10 μs is a well-known empirical value. The result obtained from convergence analysis is as
expected.
4.4.3 Performance Analysis
For the mesh refinement corresponding to the converged solution, i.e, 1 mm, the adiabatic
analysis is nearly 5 times faster in commercial software (Figure 18). The helium-cooled models,
not being able to incorporate adaptive time stepping method, are far more time consuming than
QP3.
54
5. CONCLUSIONS
The main aim of the work was to solve the thermal problem in quench using commercial FEA
tools and to compare the results with custom software in order to find out the most efficient
method of quench simulation. Modelling of the thermal phenomena in a quench comprehend
two main challenges i) the heat generation load and the material properties have non-linear
dependency on temperature and magnetic field ii) the heat transfer between the conductor and
helium goes through different transfer and boiling regimes depending upon temperatures, heat
flux and integrated heat.
In the beginning of the thesis project different approaches to quench simulation were studied in
the literature [10-16, 22-24 and 35-36]. The differential heat diffusion equation and the
empirical heat transfer equation representing the thermal problem in a quench are formulated
in Chapter 2. The practical and theoretical information about quench simulation are summarized
in Chapter 3. Two types of numerical model, adiabatic and the one including helium, were
constructed using commercial simulation environments. The adiabatic model is based on the
differential equation and solves the thermal problem in the superconductor. The helium-cooled
model is based on both the differential equation and the empirical equation, and solves the
thermal problem in a quench. Chapter 4 presents the models and the analysis carried out to
validate the results with custom software, i.e, QP3. The stability and convergence of the models
were checked for different spatial and temporal discretization. Based on the observed results
and the initial literature review we can now make the following conclusions.
The Problem arising from the non-linear material properties and temperature dependent
load was easily solved using ANSYS and COMSOL. The simulation environments included a
feature where one can create non-linear material tables and temperature-dependent loads, and
provided a straight forward solution to build the adiabatic models. The adiabatic models
estimated the hot-spot temperature and quench propagation speed with desired accuracy and
were found to be stable for a large number of elements.
Nevertheless, modelling of the heat transfer between the conductor and the helium was difficult
in both ANSYS and COMSOL. They did not provide a feature to directly implement the
empirical heat transfer equation. To define such equation a separate algorithm was constructed
using APDL scripts in ANSYS. The algorithm needs to be executed every time step to
determine the applicable regime of helium state and the adaptive time stepping scheme of the
simulation environment therefore cannot be used. Implementing a self-made time stepping
55
algorithm was beyond the scope of the thesis. A fixed time stepping scheme was selected which
resulted the helium-cooled model to be far more time consuming than QP3. The models
however estimated the hot-spot temperature and the quench propagation speed with satisfactory
accuracy.
The thermal phenomena in a quench are correctly simulated using commercial FEA tools.
Similar results are reported in [11-13], where the analysis is limited to the adiabatic scenario.
The coupled analysis presented in this thesis is a new method to take into account the non-linear
cooling behaviour of helium.
56
6. RECOMMENDATIONS FOR FUTURE WORK.
One could explore the user programmable feature in ANSYS and MatLab/COMSOL, to see if
it is possible to reconcile the regime switching algorithm with the automatic time stepping
scheme of the simulation environment.
In order to simulate quench propagation in other directions than the longitudinal, one needs
to build 2D/3D model representing a magnet coil. The 1D model constructed in ANSYS is
based on the finite element method and one can easily modify it to build 2D/3D model. Apart
from the superconductor, the conductor, and the helium, a magnet coil consists of insulation
material. One can use contact elements to represent the insulation between the cable domains
in the same way as was used to model non-linear heat transfer between the conductor and
helium; see Appendix 13.
For the simulation of large 3D models, one can explore the use of in-build adaptive spatial
discretization method. In ANSYS the method is limited to static analysis. COMSOL however
provides the feature for transient analysis as well. It is recommended to check how effectively
the method can be implemented and how much is the contribution in reducing the computation
cost.
57
7. REFERENCES
[1] P. Bauer, Stability of Superconducting Strands for Accelerator Magnets. PhD thesis,
Technical University Vienna, 1996.
[2] M. N. Wilson, Superconducting Magnets, Oxford University Press, USA, 1990.
[3] K.H. Mess, Superconducting accelerator magnets, World Scientific press, Singapore,
1996.
[4] AP. Verweij, Electrodynamics of Superconducting Cables in Accelerator Magnets, PhD
thesis, University of Twente Enschede, 1995.
[5] O. Bruning et al., LHC Design Report, Scientific Information Service, Geneva, 2004.
[6] G. Willering, Stability of Superconducting Rutherford Cables. PhD thesis, University of
Twente Enschede, 2009.
[7] P. P. Granieri, Heat Transfer between the Superconducting Cables of the LHC
Accelerator Magnets and the Superfluid Helium Bath. PhD thesis, EPFL Lausanne, 2012.
[8] A. Verweij, QP3 User’s Manual, CERN EDMS: 1150045, 2008.
[9] L. Bottura et al., THEA Thermal, Hydraulic and Electric Analysis of Superconducting
Cable. CryoSoft, Geneva, 2003.
[10] S.Russenschuck, Field Computation for Accelerator Magnets, Strauss –GmbH,
Moerlenbach, 2008.
[11] S. Caspi et al., Calculating quench propagation with ANSYS, IEEE Transactions on
Applied Superconductivity, 13(2):1714–1717, 2003.
[12] R. Yamada et al., 2D/3D quench simulation using ANSYS for epoxy impregnated NB3Sn
high field magnets, IEEE Transactions on Applied Superconductivity, 13(2):1696-1670,
2003.
[13] G. Volpini. Quench propagation in 1-d and 2-d models of high current superconductors,
Proceedings of the COMSOL Conference Milan, 2009.
[14] K.B. Kumar et al., Novel techniques to solve sets of coupled differential equations with
SPICE, IEEE Transactions on Circuits and Devices Magazine, 7(1):11-14, 1991.
[15] G. J. C. Aird et al., Coupled Transient Thermal and Electromagnetic Finite Element
Simulation of Quench in Superconducting Magnets, Proceedings of ICAP Chamonix,
2006.
[16] S.Posen et al., Copuled Electromagnetic-Thermal-Mechanical Simulations of
Superconducting RF Cavities, Proceedings of IPAC Kyoto, 2010.
58
[17] M.S. Lubell et al., Empirical scaling formulas for critical current and critical field for
commercial NbTi, IEEE Transactions on magnetics, 19(3):754-757, 1983.
[18] T.Schreiner, Current distribution inside Rutherford-type superconducting cables and
impacto on performance of LHC dipoles, PhD thesis, Technical University Viena, 2002.
[19] E.R. Bielert et al., Implementation of the superfluid helium phase transition using finite
element modelling: Simulation of transient heat transfer and He-I/He-II phase front
movement in cooling channels of superconducting magnets, ELSEVIER Cryogenics
53:78-85, 2013.
[20] S.W. Van Sciver, Helium Cryogenics, Springer, New York, 2012.
[21] Michal R.Gosz, Finite Element Application in solids, structures and heat transfer, CRC
press, London, 2006.
[22] N. Schwerg et al., Quench Simulation in an Integrated Design Environment for
Superconducting Magnets, IEEE Transactions on Magnetics, 44(6):934-937, 2008.
[23] M.Wake et al., “Complete Quench Simulation of Large Solenoid Magnet”, IEEE
Transactions on Applied Superconductivity, 22(3), 2012.
[24] V.Zermeno, Computation of Superconducting Generators for Wind Turbine
Applications, PhD thesis, Technical University of Vienna, 2012.
[25] M.A. Crisfield, Non-Linear Finite Element Analysis of Solids and Structures,
John Wiley & Sons, New York, 1997.
[26] O. C. Zienkiewicz, Finite element method, its basics and fundamentals, ELSEVIER,
USA, 2001.
[27] B. Auchmann, Numerical Treatment of Eddy Currents in LHC Dipole Magnets Using
BEM-FEM Coupling, Master Thesis, Wien Bergheideng, 2001.
[28] J.C.Sanders, Nonlinear, Transient Conduction Heat Transfer Using a Discontinuous
Galerkin Hirarchical Finite Element Method, Master Thesis, 2004.
[29] G.Noh et al., An explicit time integration scheme for the analysis of wave propagations,
ELSEVIER computer and structures, 129:178-193, 2013
[30] M. O. Domingues et al., An adaptive multiresolution scheme with local time stepping for
evolutionary PDEs, Journal of computation physics, 227:3758–3780, 2008.
[31] ANSYS Documentation, https://support.ansys.com, Last checked Dec. 2014.
[32] V.LE. Canh, Novel Numerical Procedures for limit analysis of structures, PhD thesis,
University of Sheffield, 2009.
[33] B. Auchmann, The Coupling of Discrete Electromagnetism and the Boundary Element
Method for the Simulation of Accelerator Magnets, PhD thesis, Wien Bergheideng, 2004.
59
[34] H.S.R.Rondan, Material Laws and Numerical Methods in Applied Superconductivity,
PhD thesis, University of Zaragoza Spain, 2012.
[35] V. Bansal et al., 3D Design, Electro-Thermal Simulation and Geometrical Optimization
of spiral Platinum Micro-heaters for Low Power Gas sensing applications using
COMSOL, Proceedings of the 2001 COMSOL Conference in Bangalore, 2001.
[36] D. Hu et al., 3D Modelling of All-Superconducting Synchronous Electric Machines by
the Finite Element Method, Proceedings of the 2014 COMSOL Conference in
Cambridge, 2014.
[37] K. Dahlerup Pettersen et al., The Protection System for the Superconducting Elements
of the Large Hadron Collider at CERN, Proceedings of the Particle Accelerator
Conference New York, 1999.
[38] R. Denz et al., Electronic Systems for the Protection of Superconducting Elements in the
LHC, IEEE Transactions on Applied Superconductivity, 16(2):1725-1728, 2006.
[39] E. Fornasiere et al., Status of the Activities on the Nb3Sn Dipole SMC and of the
Design of the RMC, IEEE Transactions on Applied Superconductivity, 23(3), 2013.
60
List of Figures
Figure 1 Critical surface for Nb-Ti [2]. ...................................................................................... 8
Figure 2 The Large Hadron Collider [5]. ................................................................................... 9
Figure 3 a) Symbolic representation of the LHC ring. b) Cross-section of coil of
superconducting dipole magnet. c) Rutherford-type cable. d) Cross-section of
Rutherford-type cable. e) Cross-section of a superconducting strand [1]. ....................... 10
Figure 4 Critical surface of Nb-Ti superconductor showing Tc0, Bc0, and Tcs for two different
working points along the load line [2]. ............................................................................ 16
Figure 5 Helium Phase Diagram [18] ...................................................................................... 17
Figure 6 Different assumed regimes to explain the non-linear heat transfer. .......................... 18
Figure 7 Arbitrary domain subjected to internal heat source and surface flux distribution. .... 20
Figure 8 Schematical view of domain discretization and linear solution approximation. ....... 26
Figure 9 Linear shape functions Nik, the local nodes of the element i, k = 1, 2 (4.13). ........... 27
Figure 10 Nodal shape function φi (x) at location xi. ............................................................... 27
Figure 11 Assumed geometry of the superconducting wire. .................................................... 34
Figure 12 Gaussian type temperature profile representing the initial heat pulse. .................... 35
Figure 13 Cross-section of the coil of LHC dipole magnets. Black dot indicate the location of
the strand whose cross-section properties, nominal current and magnetic field distribution
are taken as a reference for the simulation. ...................................................................... 36
Figure 14 Temperature distribution in the superconducting wire plotted for different time. The
black dots depicts advancing quench front. ..................................................................... 37
Figure 15 Quench front position as a function of time in reference to Figure 14. ................... 38
Figure 16 Upper: Temperature profile of the superconducting wire estimated by adiabatic
models of different simulation environment, Lower: Difference in Temperatures profile in
reference to QP3. .............................................................................................................. 39
Figure 17 Upper: Quench speed plotted for different mesh refinements, Lower: Change in
quench speed with respect to mesh refinement. ............................................................... 40
Figure 18 Total computation time plotted for different mesh refinements. The plotted results
corresponds to the convergence plot presented in Figure 17. Dashed lines indicates the
linear behaviour of the curves. ......................................................................................... 41
61
Figure 19 Model 1 helium cooled model. ................................................................................ 42
Figure 20 Block diagram of the algorithm that switches appropriate element every time steps.
.......................................................................................................................................... 44
Figure 21 Model 2 helium cooled model. ................................................................................ 44
Figure 22 Block diagram of the algorithm that calculates heat transfer coefficient of the contact
elements every time steps. ................................................................................................ 45
Figure 23 Model 3 Helium cooled model. ............................................................................... 46
Figure 24 Block diagram of the algorithm that updates the load values to the elements every
time steps. ......................................................................................................................... 47
Figure 25 Contour plot of Temperatures (K) obtained for from the analysis for the setting
summarised in Table 5, Uppper: From Model 2, Lower: From Model 3. ....................... 48
Figure 26 Temperature profiles of the superconducting wire, at different time, estimated by
ANSYS Model 1, for the setting summarised in the Table 5. .......................................... 49
Figure 27 Temperature profile of the superconducting wire estimated by ANSYS Models and
QP3 at time t = 0.2 s. ........................................................................................................ 50
Figure 28 Quench front position with respect to time estimated by ANSYS models and QP3.
.......................................................................................................................................... 51
Figure 29 Convergence plot: Temperature profile of the superconducting wire at t = 0.01 s for
different time step sizes. ................................................................................................... 52
Figure 30 Upper: hot-spot temperature for different time step sizes. Lower: Difference in
hot-spot temperature between two adjacent time step sizes. ............................................ 52
Figure 31 The geometry of the ANSYS model representing an SMC. Upper: 2D model, Lower
left: 3D model and Lower right: cross-section view of the 3D model. ............................ 72
62
List of Tables
Table 1 Geometry parameters used for the quench simulation of 1D adiabatic model. .......... 33
Table 2 Additional parameters used for the simulation of the 1D adiabatic model. ................ 37
Table 3 Variation in hot-spot temperature and quench speed in reference to QP3. Results
corresponds to Figure 16. ................................................................................................. 39
Table 4 Contact elements and their corresponding real constants. .......................................... 43
Table 5 Simulation parameters defined for the quench simulation in helium cooled models. 48
Table 6 Variation in the hot-spot temperature and quench speed estimated by ANSYS Models
in reference to QP3. .......................................................................................................... 51
63
APPENDICES
1. Analytical Model
To simplify the problem let us consider an adiabatic case, i.e. no helium cooling and no heat
goes in and out of the magnet coil. The relation between the total Joule heat and the temperature
rise in the coil can be expressed by time dependent heat balance equation [2 and 10].
𝜌el
𝑎cu𝐼(𝑡)2 = 𝑐p(𝑇)𝑎cable
𝑑𝑇
𝑑𝑡. (1)
Where ρel is the electrical resistivity of copper. acu is the copper cross section, acable total cross
section, I(t) the current and cp(T) the temperature dependent equivalent heat capacity of the
cable. The resistivity of Nb-Ti superconductor is orders of magnitude less than that of copper
at normal conducting state so its resistivity can be neglected. Assuming that the quench starts
at t = 0 s with initial temperature T0 and the temperature at time t is T, the separation of variables
and integration yields [10]
∫ 𝐼(𝑡)2d𝑡 = 𝑎cu𝑎cable ∫𝐶p(𝑇)
𝜌eld𝑇
𝑇
𝑇0
𝑡
0 . (2)
The term on the left hand side of the equation, the time integral of the square of the current, is
called MIIts and is usually expressed in units of 106 A2s [10]. It represents the quench load and
can be determined by measuring the current decay of the magnet during the quench. Using (2)
we can roughly estimate the average temperature of the coil during the quench (it is a worst-
case scenario).
2. Quench detection and Magnet protection
A quench is detected by measuring the voltage across the magnet. The voltage measured by
voltage taps in the magnet includes the inductive (because of the changing current in the magnet
over time) and resistive voltage. Thus, the detection system should be able to subtract the
inductive voltage from the measurement in order to obtain resistive voltage component
[37 and 38].
Magnet protection mainly consists of two methods: i) passive protection ii) active protection.
In passive protection scheme a diode or a resistor is connected to the magnet which provides
parallel path to the current. The scheme works for the magnet persisting cryogenic stabilization
even after the quench. The cryogenic stabilization means either the matrix material is sufficient
64
enough to withstand the current decay without overheating or the Ohmic heating in the copper
matrix is smaller than the rate of heat transfer to helium bath [37 and 38].
In active protection scheme once the quench is detected an appropriate protection system is
triggered. The current decay is speeded up by extracting energy into the dump resistor and/or
quench heaters are fired at several location of the magnet to speed up the quench propagation.
The stored energy in the coil is dissipated either outside of magnet or over the larger fraction
of magnet coil thus limiting the increase of the peak temperature [10].
3. Heat transfer in bulk helium
The heat transfer in the bulk helium can be explained by two regimes: laminar (Landau) regime
and turbulent (Gorter-Mellink) regime. Below a certain limiting heat flux there exists a laminar
friction-less counter-flow between the normal component and the superfluid component of
helium. The non-viscous superfluid component tends to move towards the heated surface, while
normal component moves away taking the heat along with it. Upon crossing the limit of heat
flux the relative velocity of counter-flow exceeds the critical velocity. The counter-flow being
no longer frictionless becomes turbulent [19].
4. Non-linear inductance induced losses
“For the LHC main dipoles it takes about 1200s to reach from the injection field of 0.58 T to
the nominal field of 8.4 T” [4]. The changing magnetic field with respect to time during the
up/down ramp produces coupling currents at contact resistances inside the cables and eddy
currents in the copper spacers, collars and yokes. The coupling currents are: IFCCs
(Interfilament coupling current), ISCCs (Interstrand coupling current) and BICCs
(Boundary-induced coupling currents). ISCCs induces joule heating in the contact resistance
and IFCCs induces joule heating in the copper matrix. If the heat generated raises the
temperature above the current-sharing temperature quench initiates.
65
5. Resistivity of Cu
For residual resistivity ratio, RRR = 150.
6. Critical Current for Nb-Ti Superconductor
T B( )1.7
RRR
1
2.33 109
T5
9.57 105
T3
163
T
108( )
0.37 0.0005RRR( ) B 1010( )
m( )
66
7. Heat capacity of Nb-Ti
[Jkg-1K-1]
a b c d e for T<Tc 0 4.910E+01 0 6.400E+01 0.000E+00
for Tc<T<20 K 0 1.624E+01 0 9.280E+02 0.000E+00
for 20 K<T<50 K -2.177E-01 1.198E+01 5.537E+02 -7.846E+03 4.138E+04
for 50 K<T<175 K -4.820E-03 2.976E+00 -7.163E+02 8.302E+04 -1.530E+06
for 175 K<T<500 K -6.290E-05 9.296E-02 -5.166E+01 1.3706E+04 1.240E+06
for 500 K<T<1000 K 0 0 -2.570E-01 9.555E+02 2.450E+06
For critical temperature, Tc = 8.07 K.
67
8. Heat capacity of Cu
[Jkg-1K-1]
a b c d e for T<10 K -3.080E-02 7.230E+00 -2.129E+00 1.019E+02 2.563E+00
for 10 K<T<40 K -3.045E-01 2.987E+01 -4.556E+02 3.470E+03 -8.250E+03
for 40 K<T<125 K 4.190E-02 -1.402E+01 1.509E+03 -3.160E+04 1.784E+05
for 125 K<T<300 K -8.480E-04 8.419E-01 -3.255E+02 6.059E+04 -1.290E+06
for 300 K<T<500 K -4.800E-05 9.173E-02 -6.412E+01 2.036E+04 1.030E+06
for 500 K<T<1000 K 0.000E+00 1.200E-05 -2.149E-01 1.004E+03 3.180E+06
68
9. Thermal conductivity of Cu
For magnetic field, B = 2.88 T, residual resistivity ratio, RRR = 150.
kcu T B( )2.4510
8 T
1.7
RRR
1
2.33 109
T5
9.57 105
T3
163
T
108( )
0.37 0.0005RRR( ) B 1010( )
W m1
K1
69
10. Thermal conductivity of He
[Wm-1K-1]
a b c d e f g for T<0.3 K 0 0 0 0 0 0 2.48E+03
for 0.3 K<T<8 K 2.16E-01 -6.94E+0 8.90E+01 -5.81E+2 2.05E+3 -3.79E+3 3.45E+03
for 8 K<T<20 K 0 -1.21E-3 8.81E-2 -2.56E+0 3.76E+01 -2.79E+2 1.03E+03
for 20 K<T<1000 K 2.02E-11 -2.11E-8 8.53E-06 -1.63E-3 1.22E-01 4.80E+00 1.00E+02
70
11. Heat capacity of He
T(K) Cp(J/kg/K) T Cp T Cp T Cp T Cp
1.6 1616 1.944 4457.8 2.104 7552.3 2.264 2758.9 2.424 2237.5
1.7 2210.1 1.952 4562 2.112 7826 2.272 2701.8 2.432 2228.8
1.8 2962.5 1.96 4669.2 2.12 8130.8 2.28 2650.8 2.44 2220.9
1.808 3031 1.968 4779.7 2.128 8477.1 2.288 2605.1 2.448 2213.8
1.816 3100.9 1.976 4893.6 2.136 8882.3 2.296 2564 2.456 2207.5
1.824 3172.2 1.984 5011.2 2.144 9379.4 2.304 2526.8 2.464 2201.8
1.832 3245 1.992 5132.7 2.152 10044 2.312 2493.1 2.472 2196.7
1.84 3319.3 2 5258.5 2.16 11139 2.32 2462.4 2.48 2192.3
1.848 3395.3 2.008 5388.9 2.168 7817.4 2.328 2434.5 2.488 2188.4
1.856 3472.9 2.016 5524.3 2.176 5185.5 2.336 2409 2.496 2191
1.864 3552.2 2.024 5665.1 2.184 4440 2.344 2385.7 2.5 2192.9
1.872 3633.3 2.032 5811.9 2.192 4013 2.352 2364.4 2.6 2221.2
1.88 3716.2 2.04 5965.1 2.2 3721.1 2.36 2344.8 2.7 2245.5
1.888 3801.1 2.048 6125.6 2.208 3503.7 2.368 2326.9 2.8 2292.7
1.896 3888 2.056 6294.2 2.216 3333.1 2.376 2310.5 2.9 2370.5
1.904 3977.1 2.064 6471.8 2.224 3194.5 2.384 2295.5 3 2479.3
1.912 4068.3 2.072 6659.8 2.232 3079.1 2.392 2281.7 3.1 2615.4
1.92 4161.9 2.08 6859.5 2.24 2981.2 2.4 2269.1 3.2 2763.7
1.928 4257.9 2.088 7073 2.248 2896.9 2.408 2257.6 3.3 2919.7
1.936 4356.5 2.096 7302.8 2.256 2823.5 2.416 2247.1 3.4 3081.9
100
101
102
103
0
2000
4000
6000
8000
10000
12000
Temperature, T(K)
Hea
t ca
paci
ty,
Cp(
J/K
g/K
)
71
12. Density of He
T(K) ρ(Kg/m^3) T ρ T ρ T ρ T ρ
1.6 147.25 1.76 147.38 1.92 147.59 2.08 147.95 2.24 148.23
1.608 147.2556 1.768 147.388 1.928 147.6 2.088 147.97 2.248 148.21
1.616 147.2612 1.776 147.396 1.936 147.62 2.096 148 2.256 148.18
1.624 147.2668 1.784 147.404 1.944 147.63 2.104 148.03 2.264 148.16
1.632 147.2724 1.792 147.412 1.952 147.65 2.112 148.05 2.272 148.14
1.64 147.278 1.8 147.42 1.96 147.66 2.12 148.08 2.28 148.11
1.648 147.2836 1.808 147.43 1.968 147.68 2.128 148.11 2.288 148.08
1.656 147.2892 1.816 147.44 1.976 147.69 2.136 148.15 2.296 148.06
1.664 147.2948 1.824 147.45 1.984 147.71 2.144 148.18 2.304 148.03
1.672 147.3004 1.832 147.46 1.992 147.73 2.152 148.22 2.312 148
1.68 147.306 1.84 147.47 2 147.74 2.16 148.27 2.32 147.97
1.688 147.3116 1.848 147.48 2.008 147.76 2.168 148.32 2.328 147.94
1.696 147.3172 1.856 147.49 2.016 147.78 2.176 148.33 2.336 147.91
1.704 147.324 1.864 147.51 2.024 147.8 2.184 148.33 2.344 147.88
1.712 147.332 1.872 147.52 2.032 147.82 2.192 148.32 2.352 147.85
1.72 147.34 1.88 147.53 2.04 147.84 2.2 148.31 2.36 147.82
1.728 147.348 1.888 147.54 2.048 147.86 2.208 148.3 2.368 147.79
1.736 147.356 1.896 147.55 2.056 147.88 2.216 148.28 2.376 147.75
1.744 147.364 1.904 147.57 2.064 147.9 2.224 148.26 2.384 147.72
1.752 147.372 1.912 147.58 2.072 147.93 2.232 148.25 2.392 147.69
100
101
102
103
0
50
100
150
Temperature, T(K)
Den
sity
,
(Kg/
m3)
72
13. 2D and 3D implication of 1D model
In this section we discuss the transformation of the ANSYS 1D adiabatic model to 2D/3D
models simulating quenches in a magnet coil. As a reference we consider an SMC (Short Model
Coil) designed at CERN to test new magnet technologies [39]. Figure 31 shows the ANSYS
representation of an SMC.
Figure 31 The geometry of the ANSYS model representing an SMC. Upper: 2D model, Lower left: 3D model and
Lower right: cross-section view of the 3D model.
SMC consists of 21 - 35 layers of Rutherford-type cable each made up of 14 - 40 Nb3Sn strands.
The cables are insulated from one another with a thin layer of insulation material, S2-Glass.
The upper part of Figure 31 is a cut in the x-y plane through the straight section of an
SMC. The horizontal line represents Rutherford-type cables. The cross-section area defined to
the line is equivalent to the area of copper fraction in a cable. The vertical links in between the
horizontal lines are 2 noded contact elements, “LINK33”, representing insulation between the
cables.
In the 3D model the cables are modelled using 8 noded solid elements, “SOLID70”. To
model insulation between the cables, we use contact elements, “LINK33”, in between the solid
elements as shown in lower right corner of Figure 31. The length of the coil, total number of
cables and the scale of longitudinal discretization are taken as model parameters.
The material properties, magnetic field, current and other setting assigned to the models is
in the same way as was done for 1D strand model. The APDL scripts to build 2D and 3D quench
simulation models representing the SMC can be found in Appendices 18 and 19 respectively.
X
Y
X
Y
Z
Z
Y
73
14. APDL scripts: Adiabatic model
finish
/clear
/prep7
/nerr,5
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Input Parameters !!!!!!!!!!!!!!!!!!!!!!!!!!!!!
length = 1
toper = 1.9
tpick= 20
tpickloc = 0
b = 2.88
iop = 150
acu = 0.5547e-6
dcu = 0.8404e-3
rrr = 150
fcu = 0.62263
division = 1000
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
bc20 = 14.5
tco = 9.2
c1 = 3449
c2 = -257
tc = tco*(1-b/bc20)**0.59
tcs = tc*(1-iop/(c1+c2*b))
rhocu = 8960
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Material Property tables !!!!!!!!!!!!!!!!!!!
tmax = 100
*dim,isc,table,tmax,1,1,temp
*dim,rescu,table,tmax,1,1,temp
*dim,kcu,table,tmax,1,1,temp
*dim,cpcu,table,tmax,1,1,temp
*dim,cpnbti,table,tmax,1,1,temp
*dim,cpwire,table,tmax,1,1,temp
*dim,heatgen,table,tmax,1,1,temp
!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Current in superconductor !!!!!!!!!!!!!!!!!!!!!
isc(0,1) = 0.0
*do,i,1,tmax,1
isc(i,0) = i
*if,i,lt,tcs,then
isc(i,1) = iop
*elseif,tcs,le,i,and,i,lt,tc,then
isc(i,1) = iop*(1-(i-tcs)/(tc-tcs))
*else
isc(i,1) = 0
*endif
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!! Copper Resistivity !!!!!!!!!!!!!!!!!!!!!!!!!!!
a0 = 1.7
a1 = 2.33e9
a2 = 9.57e5
a3 = 163
rescu(0,1) = 0.0
*do,i,1,tmax,1
rescu(i,0) = i
rescu(i,1) = (a0/rrr+1/(a1/i**5+a2/i**3+a3/i))*10**(-8)+(0.37+0.0005*rrr)*b*10**(-10)
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!! Copper Thermal Conductivity !!!!!!!!!!!!!!!!!!!!
kcu(0,1) = 0.0
*do,i,1,tmax,1
74
kcu(i,0) = i
kcu(i,1) = 2.45*10**(-8)*i/rescu(i,1)
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Heat Generation !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
heatgen(0,1) = 0.0
*do,i,1,tmax,1
heatgen(i,0) = i
heatgen(i,1) = rescu(i,1)*((iop-isc(i,1))**2)/acu**2
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Copper Heat Capacity !!!!!!!!!!!!!!!!!!!!!!!!!!
cpcu(0,1) = 0.0
*do,i,1,tmax,1
cpcu(i,0) = i
*if,i,lt,10,then
a = (-3.080E-02)*i**4
b = (7.230E+00)*i**3
c = (-2.129E+00)*i**2
d = (1.019E+02)*i
e = 2.563E+00
*elseif,10,le,i,and,i,lt,40,then
a = (-3.045E-01)*i**4
b = (2.987E+01)*i**3
c = (-4.556E+02)*i**2
d = (3.470E+03)*i
e = -8.250E+03
*elseif,40,le,i,and,i,lt,125,then
a = (4.190E-02)*i**4
b = (-1.402E+01)*i**3
c = (1.509E+03)*i**2
d = (-3.160E+04)*i
e = 1.784E+05
*elseif,125,le,i,and,i,lt,300,then
a = (-8.480E-04)*i**4
b = (8.419E-01)*i**3
c = (-3.255E+02)*i**2
d = (6.059E+04)*i
e = -1.290E+06
*elseif,300,le,i,and,i,lt,500,then
a = (-4.800E-05)*i**4
b = (9.173E-02)*i**3
c = (-6.412E+01)*i**2
d = (2.036E+04)*i
e = 1.030E+06
*else
a = (0.000E+00)*i**4
b = (1.200E-05)*i**3
c = (-2.149E-01)*i**2
d = (1.004E+03)*i
e = 3.180E+06
*endif
cpcu(i,1) = a+b+c+d+e
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! NbTi Heat Capacity !!!!!!!!!!!!!!!!!!!!!!!!!
cpnbti(0,dt) = 0.0
*do,i,1,tmax,1
cpnbti(i,0) = i
*if,i,lt,tc,then
a = (0.00000E+00)*i**4
b = (4.91000E+01)*i**3
75
c = (0.00000E+00)*i**2
d = (6.40000E+01)*i
e = 0.00000E+00
*elseif,tc,le,i,and,i,lt,20,then
a = (0.00000E+00)*i**4
b = (1.62400E+01)*i**3
c = (0.00000E+00)*i**2
d = (9.28000E+02)*i
e = 0.00000E+00
*elseif,20,le,i,and,i,lt,50,then
a = (-2.17700E-01)*i**4
b = (1.19838E+01)*i**3
c = (5.53710E+02)*i**2
d = (-7.84610E+03)*i
e = 4.13830E+04
*elseif,50,le,i,and,i,lt,175,then
a = (-4.82000E-03)*i**4
b = (2.97600E+00)*i**3
c = (-7.16300E+02)*i**2
d = (8.30220E+04)*i
e = -1.53000E+06
*elseif,175,le,i,and,i,lt,500,then
a = (-6.29000E-05)*i**4
b = (9.29600E-02)*i**3
c = (-5.16600E+01)*i**2
d = (1.37060E+04)*i
e = 1.24000E+06
*else
a = (0.00000E+00)*i**4
b = (0.00000E+00)*i**3
c = (-2.57000E-01)*i**2
d = (9.55500E+02)*i
e = 2.45000E+06
*endif
cpnbti(i,1) = a+b+c+d+e
*enddo
!!!!!!!!!!!!!!!!!!!!!!! Equivalent Heat Capacity !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
cpwire(0,1) = 0.0
*do,i,1,tmax,1
cpwire(i,0) = i
cpwire(i,1) = (cpcu(i,1)+(1/fcu-1)*cpnbti(i,1))/rhocu
*enddo
!!!!!!!!!!!!!!!!!!!!!!! Nodes and Elements !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
npoints = division+1
et,1,fluid116,1
r,1,dcu/2,acu
mp,dens,1,rhocu
mp,c,1,%cpwire%
mp,kxx,1,%kcu%
*do,i,1,npoints,1
n,i,(i-1)*length/division,0,0
*enddo
type,1
mat,1
real,1
*do,i,1,division,1
e,i,i+1
*enddo
Finish
!!!!!!!!!!!!!!!!!!!!!! Transient Anlysis setting !!!!!!!!!!!!!!!!!!!!!!!!!!!!
76
/config,noeldb,0
/solu
antype,4
trnopt,full
kbc,1
eqslv,sparse
bcsoption,,default
lumpm,0
time,0.182
autots,on
solcontrol,on,on
deltim,1e-5,1e-5,1e-3
neqit,1000
lnsrch,on
rescontrol,define,none,none,1
!!!!!!!!!!!!!!!!!! Initial condition, Loads, Boundary Conditions !!!!!!!!!!!!!!
*do,j,1,npoints,1
ic,j,temp,toper+(tpick-toper)*exp(-(((j-1)*length/division-tpickloc)/0.099)**2)
*enddo
nsel,all
bf,all,hgen,%heatgen%
solve
finish
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Post Analysis !!!!!!!!!!!!!!!!!!!!!!!!!!!!
/post1
nsel,s,loc,y,0
*dim,graph,table,npoints,2,1
*vget,graph(1,1),node,all,loc,x
*vget,graph(1,2),node,all,temp,y
*vplot,graph(1,1),graph(1,2)
/axlab,x,length
/axlab,y,Temperature
/replot
*cfopen,Temp_s,txt
*vwrite,graph(1,2)
(e20.5)
*cfclose
!plnsol,temp
!antime,50,0.1
Finish
77
15. APDL scripts: Helium cooled model (Model 2)
finish
/clear
/prep7
/nerr,5
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Input Parameters !!!!!!!!!!!!!!!!!!!!!!!!!!!!!
length = 1
toper = 1.9
tpick= 20
tlamda = 2.1652
tpickloc = 0
b = 2.88
iop = 150
acu = 0.5547e-6
ahe = 1.0e-6
vhe = 1.0e-6*length
ahf = 1320e-6
dcu = sqrt(4*acu/3.1416)
rrr = 150
fcu = 0.62263
division = 1000
maxtime = 0.01
nstep = 100
substep = 1
nkap = 3
akap = 200
afbii = 200
anc = 500
anb = 50000
afbi = 220
ekap = 35000
enc = 10
enb = 20000
svhe = vhe/division
sahf = ahf/division
sacu = dcu/2*3.1416*length/division
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
bc20 = 14.5
tco = 9.2
c1 = 3449
c2 = -257
tc = tco*(1-b/bc20)**0.59
tcs = tc*(1-iop/(c1+c2*b))
rhocu = 8960
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Material Property tables !!!!!!!!!!!!!!!!!!!
tmax = 100
dt = 0.1
*dim,isc,table,tmax/dt,1,1,temp
*dim,rescu,table,tmax/dt,1,1,temp
*dim,kcu,table,tmax/dt,1,1,temp
*dim,cpcu,table,tmax/dt,1,1,temp
*dim,cpnbti,table,tmax/dt,1,1,temp
*dim,cpwire,table,tmax/dt,1,1,temp
*dim,heatgen,table,tmax/dt,1,1,temp
*dim,cphe,table,tmax/dt,1,1,temp
*dim,rhohe,table,tmax/dt,1,1,temp
!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Current in superconductor !!!!!!!!!!!!!!!!!!!!!
isc(0,1) = 0.0
*do,i,1,tmax/dt,1
78
t = i*dt
isc(i,0) = t
*if,t,lt,tcs,then
isc(i,1) = iop
*elseif,tcs,le,t,and,t,lt,tc,then
isc(i,1) = iop*(1-(t-tcs)/(tc-tcs))
*else
isc(i,1) = 0
*endif
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!! Copper Resistivity !!!!!!!!!!!!!!!!!!!!!!!!!!!
a0 = 1.7
a1 = 2.33e9
a2 = 9.57e5
a3 = 163
rescu(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
rescu(i,0) = t
rescu(i,1) = (a0/rrr+1/(a1/t**5+a2/t**3+a3/t))*10**(-8)+(0.37+0.0005*rrr)*b*10**(-10)
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!! Copper Thermal Conductivity !!!!!!!!!!!!!!!!!!!!
kcu(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
kcu(i,0) = t
kcu(i,1) = 2.45*10**(-8)*t/rescu(t,1)
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Heat Generation !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
heatgen(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
heatgen(i,0) = t
heatgen(i,1) = rescu(t,1)*((iop-isc(t,1))**2)/acu**2
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Copper Heat Capacity !!!!!!!!!!!!!!!!!!!!!!!!!!
cpcu(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
cpcu(i,0) = t
*if,t,lt,10,then
a = (-3.080E-02)*t**4
b = (7.230E+00)*t**3
c = (-2.129E+00)*t**2
d = (1.019E+02)*t
e = 2.563E+00
*elseif,10,le,t,and,t,lt,40,then
a = (-3.045E-01)*t**4
b = (2.987E+01)*t**3
c = (-4.556E+02)*t**2
d = (3.470E+03)*t
e = -8.250E+03
*elseif,40,le,t,and,t,lt,125,then
a = (4.190E-02)*t**4
b = (-1.402E+01)*t**3
c = (1.509E+03)*t**2
d = (-3.160E+04)*t
e = 1.784E+05
*elseif,125,le,t,and,t,lt,300,then
a = (-8.480E-04)*t**4
79
b = (8.419E-01)*t**3
c = (-3.255E+02)*t**2
d = (6.059E+04)*t
e = -1.290E+06
*elseif,300,le,t,and,t,lt,500,then
a = (-4.800E-05)*t**4
b = (9.173E-02)*t**3
c = (-6.412E+01)*t**2
d = (2.036E+04)*t
e = 1.030E+06
*else
a = (0.000E+00)*t**4
b = (1.200E-05)*t**3
c = (-2.149E-01)*t**2
d = (1.004E+03)*t
e = 3.180E+06
*endif
cpcu(i,1) = a+b+c+d+e
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! NbTi Heat Capacity !!!!!!!!!!!!!!!!!!!!!!!!!
cpnbti(0,dt) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
cpnbti(i,0) = t
*if,t,lt,tc,then
a = (0.00000E+00)*t**4
b = (4.91000E+01)*t**3
c = (0.00000E+00)*t**2
d = (6.40000E+01)*t
e = 0.00000E+00
*elseif,tc,le,t,and,t,lt,20,then
a = (0.00000E+00)*t**4
b = (1.62400E+01)*t**3
c = (0.00000E+00)*t**2
d = (9.28000E+02)*t
e = 0.00000E+00
*elseif,20,le,t,and,t,lt,50,then
a = (-2.17700E-01)*t**4
b = (1.19838E+01)*t**3
c = (5.53710E+02)*t**2
d = (-7.84610E+03)*t
e = 4.13830E+04
*elseif,50,le,t,and,t,lt,175,then
a = (-4.82000E-03)*t**4
b = (2.97600E+00)*t**3
c = (-7.16300E+02)*t**2
d = (8.30220E+04)*t
e = -1.53000E+06
*elseif,175,le,t,and,t,lt,500,then
a = (-6.29000E-05)*t**4
b = (9.29600E-02)*t**3
c = (-5.16600E+01)*t**2
d = (1.37060E+04)*t
e = 1.24000E+06
*else
a = (0.00000E+00)*t**4
b = (0.00000E+00)*t**3
c = (-2.57000E-01)*t**2
d = (9.55500E+02)*t
e = 2.45000E+06
80
*endif
cpnbti(i,1) = a+b+c+d+e
*enddo
!!!!!!!!!!!!!!!!!!!!!!! Equivalent Heat Capacity !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
cpwire(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
cpwire(i,0) = t
cpwire(i,1) = (cpcu(t,1)+(1/fcu-1)*cpnbti(t,1))/rhocu
*enddo
!!!!!!!!!!!!!!!!!!!!!!! Helium Heat Capacity !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
*TREAD,cphe,'cphe','txt','G:\Projects\lhccm\QuSi_ANSYS\apdl_projects',0
!!!!!!!!!!!!!!!!!!!!!!!!!! Helium Density !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
*TREAD,rhohe,'rhohe','txt','G:\Projects\lhccm\QuSi_ANSYS\apdl_projects',0
!!!!!!!!!!!!!!!!!!!!!!! Nodes and Elements !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
npoints = division+1
et,1,fluid116,1
et,2,mass71,,,0
et,3,link34,,3,3
r,1,dcu,acu
r,2,svhe
r,3,sahf,0,0
r,4,sahf/2,0,0
mp,dens,1,rhocu
mp,c,1,%cpwire%
mp,kxx,1,%kcu%
mp,dens,2,%rhohe%
mp,c,2,%cphe%
*do,i,1,npoints,1
mp,hf,i+2,1e-2
*enddo
*do,j,1,2,1
*do,i,1,npoints,1
n,i+(j-1)*npoints,(i-1)*length/division,(j-1)*length/10,0
*enddo
*enddo
type,1
mat,1
real,1
*do,i,1,division,1
e,i,i+1
*enddo
type,2
mat,2
real,2
*do,i,1,npoints,1
e,i+npoints
*enddo
type,3
real,4
mat,3
e,1,1+npoints
type,3
real,3
mat,3
*do,i,2,npoints,1
mat,2+i
e,i,i+npoints
*enddo
finish
81
!!!!!!!!!!!!!!!!!!!!!! Transient Anlysis setting !!!!!!!!!!!!!!!!!!!!!!!!!!!!
/config,noeldb,0
/solu
antype,4
trnopt,full
kbc,1
eqslv,sparse
bcsoption,,default
lumpm,0
autots,off
nsubst,1
neqit,1000
lnsrch,on
rescontrol,define,none,none,1
!!!!!!!!!!!!!!!!!! Initial condition, Loads, Boundary Conditions !!!!!!!!!!!!!!
*do,j,1,npoints,1
ic,j,temp,toper+(tpick-toper)*exp(-(((j-1)*length/division-tpickloc)/0.1)**2)
*enddo
nsel,s,node,,1+npoints,2*npoints
ic,all,temp,toper
esel,s,type,,1
bfe,all,hgen,,%heatgen%
*dim,kenergy,array,npoints
*dim,tcu,array,npoints
*dim,the,array,npoints
*dim,regno,array,npoints
*dim,ncenergy,array,npoints
*dim,nbenergy,array,npoints
*do,j,1,nstep,1
time,maxtime*j/nstep
esel,all
nsel,all
solve
*vget,tcu(1),node,1,temp,val
*vget,the(1),node,1+npoints,temp,val
*do,i,1,npoints,1
*if,the(i),lt,tlamda,then
*if,regno(i),eq,3,then
kenergy(i) = abs(afbii*(tcu(i)-the(i)))
*else,
kenergy(i) = abs(akap*(tcu(i)**nkap-the(i)**nkap))
*endif
*if,kenergy(i),lt,ekap,then
regno(i) = 2
mp,hf,i+2,abs(akap*(tcu(i)**nkap-the(i)**nkap))/(abs(tcu(i)-the(i))+0.001)+0.001
*else
regno(i) = 3
mp,hf,i+2,afbii
*endif
*else
ncenergy(i) = abs(anc*(tcu(i)-the(i)))
nbenergy(i)= sign(1, tcu(i)-the(i))*abs(anb*(tcu(i)-the(i))**2.5)
*if,ncenergy(i),lt,enc,then
regno(i) = 4
mp,hf,i+2,anc
*elseif,nbenergy(i),lt,enb,then
regno(i) = 5
mp,hf,i+2,abs(anb*(tcu(i)-the(i))**1.5)+0.001
*else
regno(i) = 6
82
mp,hf,i+2,afbi
*endif
*endif
*enddo
*enddo
finish
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Post Analysis !!!!!!!!!!!!!!!!!!!!!!!!!!!!
/post1
nsel,s,loc,y,0
*dim,graph,table,npoints,2,1
*vget,graph(1,1),node,all,loc,x
*vget,graph(1,2),node,all,temp,y
*vplot,graph(1,1),graph(1,2)
/axlab,x,length
/axlab,y,Temperature
/replot
*cfopen,Temp_s,txt
*vwrite,graph(1,2)
(e20.5)
*cfclose
!plnsol,temp
!antime,50,0.1
Finish
83
16. APDL scripts: Helium cooled model (Model 1)
!!!!!!!!!!!!!!!!!!!!!!! Nodes and Elements !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
npoints = division+1
et,1,fluid116,1
et,2,link31,,,1
et,3,link34,,3,3
r,1,dcu,acu
r,2,dhe,ahe
r,3,ahf/division,0.0,0.0 !FBII/NC/FBI/ET3
r,4,ahf/division,1.5,0.0 !NB/ET3
r,5,ahf/division/2,0.0,0.0
r,6,ahf/division/2,1.5,0.0
mp,dens,1,rhocu
mp,c,1,%cpwire%
mp,kxx,1,%kcu%
mp,dens,2,%rhohe%
mp,c,2,%cphe%
mp,kxx,2,1e-6
mp,emis,3,1 !KAP/ET2/R3/M3
mp,hf,4,afbii !FBII/ET3/R4/M4
mp,hf,5,anc !NC/ET3/R4/M5
mp,hf,6,anb !NB/ET3/R5/M6
mp,hf,7,afbi !FBI/ET3/R4/M7
*do,i,1,npoints,1
n,i,(i-1)*length/division,0,0
*enddo
*do,i,1,npoints,1
n,i+npoints,(i-1)*length/division,lhecu,0
*enddo
type,1
mat,1
real,1
*do,i,1,division,1
e,i,i+1
*enddo
type,1
mat,2
real,2
*do,i,1,division,1
e,i+npoints,i+npoints+1
*enddo
type,2
mat,3
*do,i,1,npoints,1
*if,i,eq,1,then
r,6+i,ahf/division/2,ahf/division/2,1,akap
*else
r,6+i,ahf/division,ahf/division,1,akap
*endif
real,6+i
e,i,i+npoints
*enddo
type,3
real,5
mat,4
e,1,1+npoints
type,3
real,3
mat,4
84
*do,i,2,npoints,1
e,i,i+npoints
*enddo
type,3
real,5
mat,5
e,1,1+npoints
type,3
real,3
mat,5
*do,i,2,npoints,1
e,i,i+npoints
*enddo
type,3
real,6
mat,6
e,1,1+npoints
type,3
real,4
mat,6
*do,i,2,npoints,1
e,i,i+npoints
*enddo
type,3
real,5
mat,7
e,1,1+npoints
type,3
real,3
mat,7
*do,i,2,npoints,1
e,i,i+npoints
*enddo
Finish
!!!!!!!!!!!!!!!!!!!!!! Transient Anlysis setting !!!!!!!!!!!!!!!!!!!!!!!!!!!!
/config,noeldb,0
/solu
antype,4
trnopt,full
kbc,1
eqslv,sparse
bcsoption,,default
lumpm,0
autots,off
solcontrol,on
nsubst,substep
neqit,1000
lnsrch,on
rescontrol,define,none,none,1
!!!!!!!!!!!!!!!!!! Initial condition, Loads, Boundary Conditions !!!!!!!!!!!!!!
*do,j,1,npoints,1
ic,j,temp,toper+(tpick-toper)*exp(-(((j-1)*length/division-tpickloc)/0.099)**2)
*enddo
*do,i,1,npoints,1
ic,i+npoints,temp,toper
*enddo
*do,i,1,npoints,1
bf,i,hgen,%heatgen%
*enddo
*do,i,1,4*npoints,1
85
bfe,i+2*division+npoints,hgen,,0
*enddo
*dim,kenergy,array,npoints
*dim,ncenergy,array,npoints
*dim,nbenergy,array,npoints
*dim,tcu,array,npoints
*dim,the,array,npoints
*dim,regno,array,npoints
esel,s,elem,,1+npoints+2*division,2*division+5*npoints,1
ekill,all
*do,j,1,nstep,1
time,maxtime*j/nstep
esel,all
solve
*do,i,1,npoints,1
*get,tcu(i),node,i,temp
*get,the(i),node,i+npoints,temp
*enddo
esel,s,elem,,1+2*division,2*division+5*npoints,1
ekill,all
*do,i,1,npoints,1
*if,the(i),lt,tlamda,then
*if,regno(i),eq,3,then
kenergy(i)= abs(afbii*(tcu(i)-the(i)))
*else
kenergy(i)= abs(akap*(tcu(i)**nkap-the(i)**nkap))
*endif
*if,kenergy(i),lt,ekap,then
aakap = ahf/division*the(i)**(nkap-4)
fkap = ahf/division*tcu(i)**(nkap-4)
regno(i) = 2
esel,s,elem,,i+2*division
ealive,all
*if,i,eq,1,then
rmodif,6+i,1,aakap/2,fkap/2
*else
rmodif,6+i,1,aakap,fkap
*endif
*else
regno(i) = 3
esel,s,elem,,i+2*division+npoints
ealive,all
*endif
*else
ncenergy(i) = abs(anc*(tcu(i)-the(i)))
nbenergy(i)= sign(1, tcu(i)-the(i))*abs(anb*(tcu(i)-the(i))**2.5)
*if,ncenergy(i),lt,enc,then
regno(i) = 4
esel,s,elem,,i+2*division+2*npoints
ealive,all
*elseif,nbenergy(i),lt,enb,then
regno(i) = 5
esel,s,elem,,i+2*division+3*npoints
ealive,all
*else
regno(i) = 6
esel,s,elem,,i+2*division+4*npoints
ealive,all
*endif
*endif
86
*enddo
*enddo
finish
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Post Analysis !!!!!!!!!!!!!!!!!!!!!!!!!!!!
/post1
nsel,s,loc,y,0
*dim,graph,table,npoints,2,1
*vget,graph(1,1),node,all,loc,x
*vget,graph(1,2),node,all,temp,y
*vplot,graph(1,1),graph(1,2)
/axlab,x,length
/axlab,y,Temperature
/replot
*cfopen,Temp_s,txt
*vwrite,graph(1,2)
(e20.5)
*cfclose
!plnsol,temp
!antime,50,0.1
Finish
87
17. APDL scripts: Helium cooled model (Model 3)
!!!!!!!!!!!!!!!!!!!!!!! Nodes and Elements !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
npoints = division+1
et,1,fluid116,1
r,1,dcu,acu
mp,dens,1,rhocu
mp,c,1,%cpwire%
mp,kxx,1,%kcu%
*do,i,1,npoints,1
n,i,(i-1)*length/division,0,0
*enddo
type,1
mat,1
real,1
*do,i,1,division,1
e,i,i+1
*enddo
Finish
!!!!!!!!!!!!!!!!!!!!!! Transient Anlysis setting !!!!!!!!!!!!!!!!!!!!!!!!!!!!
/config,noeldb,0
/solu
antype,4
trnopt,full
kbc,1
autots,off
nsubst,1
eqslv,sparse
bcsoption,,default
lumpm,0
neqit,1000
lnsrch,on
rescontrol,define,none,none,1
!!!!!!!!!!!!!!!!!! Initial condition, Loads, Boundary Conditions !!!!!!!!!!!!!!
*do,j,1,npoints,1
ic,j,temp,toper+(tpick-toper)*exp(-(((j-1)*length/division-tpickloc)/0.099)**2)
*enddo
*dim,kenergy,array,npoints
*dim,ncenergy,array,npoints
*dim,nbenergy,array,npoints
*dim,fbenergy,array,npoints
*dim,tcu,array,npoints
*dim,the,array,npoints
*dim,regno,array,npoints
*do,i,1,npoints,1
the(i) = toper
*enddo
*do,j,1,2868,1
time,timestep(j,1)
esel,all
solve
*do,i,1,npoints-1,1
*get,tcu(i),node,i,temp
*if,regno(i),eq,4,then
the(i) = the(i)+ncenergy(i)*sahf/(svhe*rhohe(the(i),1)*cphe(the(i),1))*(timestep(j,1)-timestep(j-1,1))
*elseif,regno(i),eq,5,then
the(i) = the(i)+nbenergy(i)*sahf/(svhe*rhohe(the(i),1)*cphe(the(i),1))*(timestep(j,1)-timestep(j-1,1))
*elseif,regno(i),eq,6,then
the(i) = the(i)+fbenergy(i)*sahf/(svhe*rhohe(the(i),1)*cphe(the(i),1))*(timestep(j,1)-timestep(j-1,1))
*else
88
the(i) = the(i)+kenergy(i)*sahf/(svhe*rhohe(the(i),1)*cphe(the(i),1))*(timestep(j,1)-timestep(j-1,1))
*endif
*enddo
*do,i,1,npoints-1,1
*if,the(i),lt,tlamda,then
*if,regno(i),eq,3,then
kenergy(i)= abs(afbii*(tcu(i)-the(i)))
*else
kenergy(i)= abs(akap*(tcu(i)**nkap-the(i)**nkap))
*endif
bfe,i,hgen,,heatgen(tcu(i))-kenergy(i)*sahf/(acu*length/division)
*if,kenergy(i),lt,ekap,then
regno(i) = 2
*else
regno(i) = 3
*endif
*else
ncenergy(i) = abs(anc*(tcu(i)-the(i)))
nbenergy(i)= sign(1, tcu(i)-the(i))*abs(anb*(tcu(i)-the(i))**2.5)
*if,ncenergy(i),lt,enc,then
regno(i) = 4
bfe,i,hgen,,heatgen(tcu(i))-ncenergy(i)*sahf/(acu*length/division)
*elseif,nbenergy(i),lt,enb,then
regno(i) = 5
bfe,i,hgen,,heatgen(tcu(i))-nbenergy(i)*sahf/(acu*length/division)
*else
regno(i) = 6
fbenergy(i) = afbi*(tcu(i)-the(i))
bfe,i,hgen,,heatgen(tcu(i))-fbenergy(i)*sahf/(acu*length/division)
*endif
*endif
*enddo
*enddo
finish
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Post Analysis !!!!!!!!!!!!!!!!!!!!!!!!!!!!
/post1
nsel,s,loc,y,0
*dim,graph,table,npoints,2,1
*vget,graph(1,1),node,all,loc,x
*vget,graph(1,2),node,all,temp,y
*vplot,graph(1,1),graph(1,2)
/axlab,x,length
/axlab,y,Temperature
/replot
*cfopen,Temp_s,txt
*vwrite,graph(1,2)
(e20.5)
*cfclose
!plnsol,temp
!antime,50,0.1
Finish
89
18. APDL scripts: 2D quench simulation model of a magnet coil
finish
/clear,all
/prep7
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Input Values !!!!!!!!!!!!!!!!!!!!!!!!!!!!
length = 1
ruthno = 5
ttf = 2
lhecu = 240*10**(-(6-ttf))
hcu = 15e-3
wcu = 2e-3
division = 100
maxtime = 0.2
nstep = 1
substep = 1000
toper = 1.9
tpick = 20
tpickloc = 0
b = 2.88
iop = 150
rrr = 150
fcu = 0.62263
acu = 0.5547e-6
dcu = 0.8404e-3
hcu = dcu*14
wcu = dcu*2
ahf = hcu*length/(division+1)
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Derived Parameters !!!!!!!!!!!!!!!!!!!!!!!!!!
bc20 = 14.5
tco = 9.2
c1 = 3449
c2 = -257
tc = tco*(1-b/bc20)**0.59
tcs = tc*(1-iop/(c1+c2*b))
rhocu = 8960
rhokap = 1420
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Material Property tables !!!!!!!!!!!!!!!!!!!
tmax = 100
dt = 0.1
*dim,isc,table,tmax/dt,1,1,temp
*dim,rescu,table,tmax/dt,1,1,temp
*dim,kcu,table,tmax/dt,1,1,temp
*dim,cpcu,table,tmax/dt,1,1,temp
*dim,cpnb3sn,table,tmax/dt,1,1,temp
*dim,cpwire,table,tmax/dt,1,1,temp
*dim,kg10pll,table,tmax/dt,1,1,temp
*dim,kg10nl,table,tmax/dt,1,1,temp
*dim,cpg10,table,tmax/dt,1,1,temp
*dim,heatgen,table,tmax/dt,1,1,temp
!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Current in superconductor !!!!!!!!!!!!!!!!!!!!!
isc(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
isc(i,0) = t
*if,t,lt,tcs,then
isc(i,1) = iop
*elseif,tcs,le,t,and,t,lt,tc,then
isc(i,1) = iop*(1-(t-tcs)/(tc-tcs))
*else
90
isc(i,1) = 0
*endif
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!! Copper Resistivity !!!!!!!!!!!!!!!!!!!!!!!!!!!
a0 = 1.7
a1 = 2.33e9
a2 = 9.57e5
a3 = 163
rescu(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
rescu(i,0) = t
rescu(i,1) = (a0/rrr+1/(a1/t**5+a2/t**3+a3/t))*10**(-8)+(0.37+0.0005*rrr)*b*10**(-10)
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!! Copper Thermal Conductivity !!!!!!!!!!!!!!!!!!!!
kcu(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
kcu(i,0) = t
kcu(i,1) = 2.45*10**(-8)*t/rescu(t,1)
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Heat Generation !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
heatgen(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
heatgen(i,0) = t
heatgen(i,1) = rescu(t,1)*((iop-isc(t,1))**2)/acu**2
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Copper Heat Capacity !!!!!!!!!!!!!!!!!!!!!!!!!!
cpcu(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
cpcu(i,0) = t
*if,t,lt,10,then
a = (-3.080E-02)*t**4
b = (7.230E+00)*t**3
c = (-2.129E+00)*t**2
d = (1.019E+02)*t
e = 2.563E+00
*elseif,10,le,t,and,t,lt,40,then
a = (-3.045E-01)*t**4
b = (2.987E+01)*t**3
c = (-4.556E+02)*t**2
d = (3.470E+03)*t
e = -8.250E+03
*elseif,40,le,t,and,t,lt,125,then
a = (4.190E-02)*t**4
b = (-1.402E+01)*t**3
c = (1.509E+03)*t**2
d = (-3.160E+04)*t
e = 1.784E+05
*elseif,125,le,t,and,t,lt,300,then
a = (-8.480E-04)*t**4
b = (8.419E-01)*t**3
c = (-3.255E+02)*t**2
d = (6.059E+04)*t
e = -1.290E+06
*elseif,300,le,t,and,t,lt,500,then
a = (-4.800E-05)*t**4
91
b = (9.173E-02)*t**3
c = (-6.412E+01)*t**2
d = (2.036E+04)*t
e = 1.030E+06
*else
a = (0.000E+00)*t**4
b = (1.200E-05)*t**3
c = (-2.149E-01)*t**2
d = (1.004E+03)*t
e = 3.180E+06
*endif
cpcu(i,1) = a+b+c+d+e
*enddo
!!!!!!!!!!!!!!!!!!!!!!! NB3Sn Heat Capacity !!!!!!!!!!!!!!!!!!!!!!!!!
cpnb3sn(0,dt) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
cpnb3sn(i,0) = t
*if,t,lt,tc,then
a = (0.00000E+00)*t**4
b = (38.8 - 1.8*bh+0.0634*bh**2)*t**3
c = (-110*10**(-0.434*bh))*t**2
d = (207-3.83*bh+2.86*bh**2)*t
e = 0.00000E+00
*elseif,tc,le,t,and,t,lt,26.113,then
a = (0.00000E+00)*t**4
b = (7.4200E+00)*t**3
c = (0.00000E+00)*t**2
d = (1.522000E+03)*t
e = 0.00000E+00
*elseif,26.113,le,t,and,t,lt,169.416,then
a = (0.00000E+00)*t**4
b = (0.00000E+00)*t**3
c = (-6.16350E+01)*t**2
d = (1.9902E+04)*t
e = -3.058070E+05
*elseif,169.416,le,t,and,t,lt,300,then
a = (0.00000E+00)*t**4
b = (0.00000E+00)*t**3
c = (-7.46360E+00)*t**2
d = (4.411E+03)*t
e = 7.638010E+05
*else
a = (0.00000E+00)*t**4
b = (0.00000E+00)*t**3
c = (0.000E+00)*t**2
d = (0.0000E+00)*t
e = 1.415377E+06
*endif
cpnb3sn(i,1) = a+b+c+d+e
*enddo
!!!!!!!!!!!!!!!!!!!!!!! Equivalent Heat Capacity 2 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
cpwire(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
cpwire(i,0) = t
cpwire(i,1) = (cpcu(t,1)+(1/fcu2-1)*cpnb3sn(t,1))/rhocu
*enddo
!!!!!!!!!!!!!!!!!!!!!! G10 Normal Thermal Conductivity !!!!!!!!!!!!!!!!!!!!!!!!!!!!!
kg10nl(0,1) = 0.0
92
*do,i,1,tmax/dt,1
t = i*dt
kg10nl(i,0) = t
a = -4.1236
b = 13.788*log10(t)
c = -26.068*log10(t)**2
d = 26.272*log10(t)**3
e = -14.663*log10(t)**4
f = 4.4954*log10(t)**5
g = -0.6975*log10(t)**6
h = 0.0397*log10(t)**7
kg10nl(i,1) = 10**(a+b+c+d+e+f+g+h)
*enddo
!!!!!!!!!!!!!!!!!!!!!! G10 Parallel Thermal Conductivity !!!!!!!!!!!!!!!!!!!!!!!!!!!!!
kg10pll(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
kg10pll(i,0) = t
a = -2.64827
b = 8.80228*log10(t)
c = -24.8998*log10(t)**2
d = 41.1625*log10(t)**3
e = -39.8754*log10(t)**4
f = 23.1778*log10(t)**5
g = -7.95635*log10(t)**6
h = 1.48806*log10(t)**7
h1 = -0.11701*log10(t)**8
kg10pll(i,1) = 10**(a+b+c+d+e+f+g+h)
*enddo
!!!!!!!!!!!!!!!!!!!!!! G10 Heat Capacity !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
cpg10(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
cpg10(i,0) = t
a = -2.4083
b = 7.6006*log10(t)
c = -8.2982*log10(t)**2
d = 7.3301*log10(t)**3
e = -4.2386*log10(t)**4
f = 1.4294*log10(t)**5
g = -0.24396*log10(t)**6
h = 0.015236*log10(t)**7
cpg10(i,1) = 10**(a+b+c+d+e+f+g+h)
*enddo
!!!!!!!!!!!!!!!!!!!!!! Element types and Material types !!!!!!!!!!!!!!!!!!
npoints = division + 1
et,1,fluid116,1
et,2,link33
r,1,,acu*28
r,2,ahf
mp,dens,1,rhocu
mp,dens,2,rhokap
mptgen,1,90,dt,dt
mptgen,91,10,10,10
*do,i,1,100,1
*if,i,le,90,then
mpdata,kxx,1,i,kcu(i*dt,1)
*else
mpdata,kxx,1,i,kcu((i-90)*10,1)
*endif
93
*enddo
*do,i,1,100,1
*if,i,le,90,then
mpdata,c,1,i,cpwire(i*dt,1)
*else
mpdata,c,1,i,cpwire((i-90)*10,1)
*endif
*enddo
*do,i,1,100,1
*if,i,le,90,then
mpdata,kxx,2,i,kg10n1(i*dt,1)*10**ttf
*else
mpdata,kxx,2,i,kg10nl((i-90)*10,1)*10**ttf
*endif
*enddo
*do,i,1,100,1
*if,i,le,90,then
mpdata,c,2,i,cpg10(i*dt,1)*10**(-ttf)
*else
mpdata,c,2,i,cpg10((i-90)*10,1)*10**(-ttf)
*endif
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Nodes and Elements !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
*do,j,1,ruthno,1
*do,i,1,npoints,1
n,i+(j-1)*npoints,(i-1)*length/division,(j-1)*lhecu,0
*enddo
*enddo
type,1
mat,1
real,1
*do,j,1,ruthno,1
*do,i,1,division,1
e,i+(j-1)*npoints,i+(j-1)*npoints+1
*enddo
*enddo
type,2
mat,2
real,2
*do,j,1,ruthno-1,1
*do,i,1,npoints,1
e,i+(j-1)*npoints,i+j*npoints
*enddo
*enddo
finish
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Transient Analysis Setting !!!!!!!!!!!!!!!!!!!!!
/solu
antype,4
trnopt,full
kbc,0
eqslv,sparse
bcsoption,,default
lumpm,0
autots,on
time,maxtime
nsubst,substep
neqit,1000
lnsrch,on
outres,all,all
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Initial Condition, Loads, Boundary Conditions
94
*do,j,1,npoints,1
ic,j,temp,toper+(tpick-toper)*exp(-(((j-1)*length/division-tpickloc)/0.0143)**2)
*enddo
nsel,s,node,,1+npoints,6*npoints,1
ic,all,temp,toper
nsel,all
bf,all,hgen,%heatgen%
esel,s,type,,2
bfe,all,hgen,,0
esel,all
solve
finish
/post1
nsel,s,loc,y,0
*dim,graph,table,6*npoints,2,1
*vget,graph(1,1),node,all,loc,x
*vget,graph(1,2),node,all,temp,y
*vplot,graph(1,1),graph(1,2)
/axlab,x,length
/axlab,y,Temperature
/replot
*cfopen,Temp_s,txt
*vwrite,graph(1,2)
(e20.5)
*cfclose
!plnsol,temp
!antime,50,0.1
!finish
95
19. APDL scripts: 3D quench simulation model of a magnet coil
finish
/clear,all
/prep7
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Input Values !!!!!!!!!!!!!!!!!!!!!!!!!!!!
length = 1
lhecu = 240e-6
division = 100
ruthno = 5
maxtime = 0.2
toper = 1.9
tpick = 20
tpickloc = 0
bh = 2.88
iop = 150
rrr = 150
fcu = 0.62263
fcu2 = 1.25/2.25
acu = 0.5547e-6
dcu = 0.8404e-3
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Derived parameters !!!!!!!!!!!!!!!!!!!!!!!
ahf = 14*sqrt(acu)*length/(15*(division+1))
bc20 = 14.5
tco = 9.2
c1 = 3449
c2 = -257
tc = tco*(1-bh/bc20)**0.59
tcs = tc*(1-iop/(c1+c2*b))
rhocu = 8960
rhokap = 1420
rhog10 = 1760
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Material Property tables !!!!!!!!!!!!!!!!!!!
tmax = 100
dt = 0.1
*dim,isc,table,tmax/dt,1,1,temp
*dim,rescu,table,tmax/dt,1,1,temp
*dim,kcu,table,tmax/dt,1,1,temp
*dim,cpcu,table,tmax/dt,1,1,temp
*dim,cpnbti,table,tmax/dt,1,1,temp
*dim,cpwire,table,tmax/dt,1,1,temp
*dim,kkap,table,tmax/dt,1,1,temp
*dim,cpkap,table,tmax/dt,1,1,temp
*dim,cpnb3sn,table,tmax/dt,1,1,temp
*dim,cpwire2,table,tmax/dt,1,1,temp
*dim,cpg10,table,tmax/dt,1,1,temp
*dim,kg10nl,table,tmax/dt,1,1,temp
*dim,kg10pll,table,tmax/dt,1,1,temp
*dim,heatgen,table,tmax/dt,1,1,temp
!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Current in superconductor !!!!!!!!!!!!!!!!!!!!!
isc(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
isc(i,0) = t
*if,t,lt,tcs,then
isc(i,1) = iop
*elseif,tcs,le,t,and,t,lt,tc,then
isc(i,1) = iop*(1-(t-tcs)/(tc-tcs))
*else
isc(i,1) = 0
96
*endif
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!! Copper Resistivity !!!!!!!!!!!!!!!!!!!!!!!!!!!
a0 = 1.7
a1 = 2.33e9
a2 = 9.57e5
a3 = 163
rescu(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
rescu(i,0) = t
rescu(i,1) = (a0/rrr+1/(a1/t**5+a2/t**3+a3/t))*10**(-8)+(0.37+0.0005*rrr)*bh*10**(-10)
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!! Copper Thermal Conductivity !!!!!!!!!!!!!!!!!!!!
kcu(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
kcu(i,0) = t
kcu(i,1) = 2.45*10**(-8)*t/rescu(t,1)
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Heat Generation !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
heatgen(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
heatgen(i,0) = t
heatgen(i,1) = rescu(t,1)*((iop-isc(t,1))**2)/acu**2
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Copper Heat Capacity !!!!!!!!!!!!!!!!!!!!!!!!!!
cpcu(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
cpcu(i,0) = t
*if,t,lt,10,then
a = (-3.080E-02)*t**4
b = (7.230E+00)*t**3
c = (-2.129E+00)*t**2
d = (1.019E+02)*t
e = 2.563E+00
*elseif,10,le,t,and,t,lt,40,then
a = (-3.045E-01)*t**4
b = (2.987E+01)*t**3
c = (-4.556E+02)*t**2
d = (3.470E+03)*t
e = -8.250E+03
*elseif,40,le,t,and,t,lt,125,then
a = (4.190E-02)*t**4
b = (-1.402E+01)*t**3
c = (1.509E+03)*t**2
d = (-3.160E+04)*t
e = 1.784E+05
*elseif,125,le,t,and,t,lt,300,then
a = (-8.480E-04)*t**4
b = (8.419E-01)*t**3
c = (-3.255E+02)*t**2
d = (6.059E+04)*t
e = -1.290E+06
*elseif,300,le,t,and,t,lt,500,then
a = (-4.800E-05)*t**4
b = (9.173E-02)*t**3
c = (-6.412E+01)*t**2
97
d = (2.036E+04)*t
e = 1.030E+06
*else
a = (0.000E+00)*t**4
b = (1.200E-05)*t**3
c = (-2.149E-01)*t**2
d = (1.004E+03)*t
e = 3.180E+06
*endif
cpcu(i,1) = a+b+c+d+e
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! NbTi Heat Capacity !!!!!!!!!!!!!!!!!!!!!!!!!
cpnbti(0,dt) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
cpnbti(i,0) = t
*if,t,lt,tc,then
a = (0.00000E+00)*t**4
b = (4.91000E+01)*t**3
c = (0.00000E+00)*t**2
d = (6.40000E+01)*t
e = 0.00000E+00
*elseif,tc,le,t,and,t,lt,20,then
a = (0.00000E+00)*t**4
b = (1.62400E+01)*t**3
c = (0.00000E+00)*t**2
d = (9.28000E+02)*t
e = 0.00000E+00
*elseif,20,le,t,and,t,lt,50,then
a = (-2.17700E-01)*t**4
b = (1.19838E+01)*t**3
c = (5.53710E+02)*t**2
d = (-7.84610E+03)*t
e = 4.13830E+04
*elseif,50,le,t,and,t,lt,175,then
a = (-4.82000E-03)*t**4
b = (2.97600E+00)*t**3
c = (-7.16300E+02)*t**2
d = (8.30220E+04)*t
e = -1.53000E+06
*elseif,175,le,t,and,t,lt,500,then
a = (-6.29000E-05)*t**4
b = (9.29600E-02)*t**3
c = (-5.16600E+01)*t**2
d = (1.37060E+04)*t
e = 1.24000E+06
*else
a = (0.00000E+00)*t**4
b = (0.00000E+00)*t**3
c = (-2.57000E-01)*t**2
d = (9.55500E+02)*t
e = 2.45000E+06
*endif
cpnbti(i,1) = a+b+c+d+e
*enddo
!!!!!!!!!!!!!!!!!!!!!!! NB3Sn Heat Capacity !!!!!!!!!!!!!!!!!!!!!!!!!
cpnb3sn(0,dt) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
cpnb3sn(i,0) = t
98
*if,t,lt,tc,then
a = (0.00000E+00)*t**4
b = (38.8 - 1.8*bh+0.0634*bh**2)*t**3
c = (-110*10**(-0.434*bh))*t**2
d = (207-3.83*bh+2.86*bh**2)*t
e = 0.00000E+00
*elseif,tc,le,t,and,t,lt,26.113,then
a = (0.00000E+00)*t**4
b = (7.4200E+00)*t**3
c = (0.00000E+00)*t**2
d = (1.522000E+03)*t
e = 0.00000E+00
*elseif,26.113,le,t,and,t,lt,169.416,then
a = (0.00000E+00)*t**4
b = (0.00000E+00)*t**3
c = (-6.16350E+01)*t**2
d = (1.9902E+04)*t
e = -3.058070E+05
*elseif,169.416,le,t,and,t,lt,300,then
a = (0.00000E+00)*t**4
b = (0.00000E+00)*t**3
c = (-7.46360E+00)*t**2
d = (4.411E+03)*t
e = 7.638010E+05
*else
a = (0.00000E+00)*t**4
b = (0.00000E+00)*t**3
c = (0.000E+00)*t**2
d = (0.0000E+00)*t
e = 1.415377E+06
*endif
cpnb3sn(i,1) = a+b+c+d+e
*enddo
!!!!!!!!!!!!!!!!!!!!!!! Equivalent Heat Capacity 2 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
cpwire(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
cpwire2(i,0) = t
cpwire(i,1) = (cpcu(t,1)+(1/fcu2-1)*cpnb3sn(t,1))/rhocu
*enddo
!!!!!!!!!!!!!!!!!!!!!! G10 Normal Thermal Conductivity !!!!!!!!!!!!!!!!!!!!!!!!!!!!!
kg10nl(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
kg10nl(i,0) = t
a = -4.1236
b = 13.788*log10(t)
c = -26.068*log10(t)**2
d = 26.272*log10(t)**3
e = -14.663*log10(t)**4
f = 4.4954*log10(t)**5
g = -0.6975*log10(t)**6
h = 0.0397*log10(t)**7
kg10nl(i,1) = 10**(a+b+c+d+e+f+g+h)
*enddo
!!!!!!!!!!!!!!!!!!!!!! G10 Parallel Thermal Conductivity !!!!!!!!!!!!!!!!!!!!!!!!!!!!!
kg10pll(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
kg10pll(i,0) = t
99
a = -2.64827
b = 8.80228*log10(t)
c = -24.8998*log10(t)**2
d = 41.1625*log10(t)**3
e = -39.8754*log10(t)**4
f = 23.1778*log10(t)**5
g = -7.95635*log10(t)**6
h = 1.48806*log10(t)**7
h1 = -0.11701*log10(t)**8
kg10pll(i,1) = 10**(a+b+c+d+e+f+g+h)
*enddo
!!!!!!!!!!!!!!!!!!!!!! G10 Heat Capacity !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
cpg10(0,1) = 0.0
*do,i,1,tmax/dt,1
t = i*dt
cpg10(i,0) = t
a = -2.4083
b = 7.6006*log10(t)
c = -8.2982*log10(t)**2
d = 7.3301*log10(t)**3
e = -4.2386*log10(t)**4
f = 1.4294*log10(t)**5
g = -0.24396*log10(t)**6
h = 0.015236*log10(t)**7
cpg10(i,1) = 10**(a+b+c+d+e+f+g+h)
*enddo
!!!!!!!!!!!!!!!!!!!!!! Element types and Material types !!!!!!!!!!!!!!!!!!!!!!!!
npoints = division + 1
et,1,solid70
et,2,link33
mp,dens,1,rhocu
mp,dens,2,rhokap
mptgen,1,90,dt,dt
mptgen,91,10,10,10
!!!!!!!!!!!!!!!!!!!!!!!!!!1 Nonlinear Material Property definition !!!!!!!!!!!!1
*do,i,1,100,1
*if,i,le,90,then
mpdata,kxx,1,i,kcu(i*dt,1)
*else
mpdata,kxx,1,i,kcu((i-90)*10,1)
*endif
*enddo
*do,i,1,100,1
*if,i,le,90,then
mpdata,c,1,i,cpwire(i*dt,1)
*else
mpdata,c,1,i,cpwire((i-90)*10,1)
*endif
*enddo
*do,i,1,100,1
*if,i,le,90,then
mpdata,kxx,2,i,kg10(i*dt,1)
*else
mpdata,kxx,2,i,kg10((i-90)*10,1)
*endif
*enddo
*do,i,1,100,1
*if,i,le,90,then
mpdata,c,2,i,cpg10(i*dt,1)
*else
100
mpdata,c,2,i,cpg10((i-90)*10,1)
*endif
*enddo
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Nodes and Elements !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
n,1, 0 ,0 ,0
n,2, length/division,0 ,0
n,3, length/division,sqrt(acu),0
n,4, 0 ,sqrt(acu),0
ngen,3,4,1,4,1,0,0,sqrt(acu)
type,1 ! Elements type for superconductor
mat,1
e,1,2,3,4,5,6,7,8
e,5,6,7,8,9,10,11,12
esel,all
egen,ruthno,12,all,,,,,,,,0,0,2*sqrt(acu)+lhecu
type,2 ! Element type for Insulation
mat,2
r,1,ahf
r,2,ahf
r,3,ahf
r,4,ahf
!!!!!!!!!!!!!!! Generating insulation elements between the superconductor element
*do,i,1,ruthno-1,1
*do,j,1,4,1
real,j
e,j+8+(i-1)*12,j+(i-1)*12+12
*enddo
*enddo
esel,s,type,,1
esel,a,real,,3,4
egen,14,12*ruthno,all,,,,,,,,0,sqrt(acu),0
esel,s,type,,1
esel,a,real,,2,3
egen,division,14*12*ruthno,all,,,,,,,,length/division,0,0
nummrg,node
finish
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Transient Analysis Setting !!!!!!!!!!!!!!!!!!!!!
/solu
antype,4
trnopt,full
kbc,0
eqslv,sparse
bcsoption,,default
lumpm,0
autots,on
time,maxtime
deltime, 1e-5, 1e-5, 5e-4
neqit,1000
lnsrch,on
outres,all,all
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Initial Condition, Loads, Boundary Conditions !!!!!
! Initial temperature profile
*do,i,1,npoints,1
nsel,s,loc,x,(i-1)*length/(division)
nsel,r,loc,y,0,5*sqrt(acu),sqrt(acu)
nsel,r,loc,z,0,2*sqrt(acu),sqrt(acu)
ic,all,temp,toper+(tpick-toper)*exp(-(((i-1)*length/division-tpickloc)/0.0141)**2)
*enddo
nsel,s,loc,z,0,2*sqrt(acu),sqrt(acu)
nsel,r,loc,y,0,5*sqrt(acu),sqrt(acu)
101
nsel,inve
ic,all,temp,toper
!Applying heat generation
esel,s,type,,1
bfe,all,hgen,,%heatgen%
esel,s,type,,2
bfe,all,hgen,,0
esel,all
nsel,all
solve
finish
/post1
!nsel,s,loc,y,0
!*dim,graph,table,6*npoints,2,1
!*vget,graph(1,1),node,all,loc,x
!*vget,graph(1,2),node,all,temp,y
!*vplot,graph(1,1),graph(1,2)
!/axlab,x,length
!/axlab,y,Temperature
!/replot
!*cfopen,Temp_s,txt
!*vwrite,graph(1,2)
!(e20.5)
!*cfclose
plnsol,temp
/contour,all,9,1.9,,tc ! The grey color represent temperature above tc
/color,smax,14
/color,outl,0
!antime,50,0.1
!finish
